A Narrative Review on Key Values Indicators of Millimeter Wave Radars for Ambient Assisted Living

Abstract

The demographic shift toward an aging population calls for innovative strategies to ensure independence, health, and quality of life in later years. In this context, Ambient Assisted Living (AAL) solutions, supported by Information and Communication Technologies (ICTs), offer promising advances for non-invasive and continuous support. Commonly, ICTs are evaluated only from the perspectives related to key performance indicators (KPIs); nevertheless, the design and implementation of such technologies must account for important psychological, social, and ethical dimensions. Radar-based sensing systems are emerging as an option due to their unobtrusive nature and capacity to operate without direct user interaction. This work explores how radar technologies, particularly those operating in the millimeter wave (mmWave) spectrum, can provide core key value indicators (KVIs) essential to aging societies, such as human dignity, trustworthiness, fairness, and sustainability. Through a review of key application domains, the paper illustrates the practical contributions of mmWave radar in Ambient Assisting Living (AAL) contexts, underlining how its technical attributes align with the complex needs of elderly care environments and produce value for society. This work uniquely integrates key value indicator (KVI) frameworks with mmWave radar capabilities to address unmet ethical needs in the AAL domain. It advances existing literature by proposing a value-driven design approach that directly informs technical specifications, enabling the alignment of engineering choices with socially relevant values and supporting the development of technologies for a more inclusive and ethical society.

1. Introduction

A major increase in life expectancy and a decline in birth rates have produced a growth in the number of people over 65 in the last few years [1,2]. Consequently, it is crucial to assist elders in maintaining a fulfilling quality of life while preventing loss of independence. Different disciplines are dealing with the issues of an aging society, including sociology, psychology, philosophy, and law [3]. Information and Communication Technology (ICT) are considered enabling technologies in the field of electrical and biomedical engineering to aid the elderly in their daily activities [4]. A recent trend involves the use of remote sensors due to the reluctance of older people to wear sensors and their low cooperation. The integration of sensors in indoor environments is referred to as AAL and permits non-intrusive monitoring of the subject. Contactless sensing can be implemented using various modalities, among which mmWave radar is frequently advocated as a privacy-preserving alternative to conventional RGB camera-based systems [5], due to its inability to capture high-resolution visual content such as facial features or environmental details, thereby substantially reducing privacy-related risks.

In the scientific literature there are some reviews, surveys and systematic literature reviews (SLRs) that deal with deep studies about the AAL environment or signal processing feature used to extract and manipulate data to achieve the results, as shown in

Table A1. Table for recent prior reviews, survey and systematic literature review.

Ref.	Summary	Ethical Issue	AAL	KVI
[10]	Radar sensors are explored as solutions for contactless analysis of vital signs. The analysed papers focus on the regression and classification of cardiac and breathing waveforms by employing state-of-the-art machine learning techniques. The metrics, datasets, and algorithms are analyzed and categorized.	Not addressed	NO	NO
[35]	The paper reviews radar-based continuous human activity recognition advancements. Various radar data representations improve classification accuracy. Future research should focus on multiple-subject and multi-activity classification.	Not addressed in the paper.	YES	NO
[49]	The paper reviews AAL systems supported by IoT technologies. The challenges and opportunities in IoT-supported AAL are outlined. The review is based on 61 papers from 643 initial results.	The paper discusses the management of data security and privacy in systems related to ambient assisted living and the Internet of Things. While privacy is treated in the paper, radar technology’s privacy implications are not discussed.	YES	NO
[46]	Radar is used to aid diagnosis and track disease progression. Ethical considerations are crucial for future applications. The paper encourages further research in this promising field.	The paper emphasizes the importance of developing ethical criteria that are context-sensitive and evaluable, which relates to usability and accuracy in the application of these technologies.	YES	NO
[50]	Heart and respiratory rates are crucial for assessing human body status. Various methods for vital signal extraction are reviewed for accuracy and efficiency. Wavelet analysis and mode decomposition enhance vital signal measurement opportunities.	Not addressed.	NO	NO
[5]	The paper explores interdisciplinary perspectives on AAL systems. It emphasizes privacy, performance, and usability in AAL technologies. The research identifies recent advancements and future research areas in AAL.	The paper stresses privacy in IoT and AAL systems, emphasizing trust, fairness, and the risks of bias in AI. It also highlights the need for usability, accuracy, and economic sustainability in decision support systems.	YES	NO
[51]	The study examines wearable tech adoption among older adults with chronic illnesses, highlighting that social influence and personalization boost engagement, while health consciousness had little impact.	The research stresses privacy concerns around wearable sensors, especially for data confidentiality. It also highlights trust, usability, and economic factors like cost-benefit in adoption.	YES	NO

in Appendix A.

Several recent studies have explored the application of mmWave radar AAL environments, including considerations of these ethical aspects. In [6], the author addresses the issue of elderly fall detection by presenting a comparative overview of existing monitoring systems, both contactless and wearable, and highlights the privacy-preserving advantages of radar technology. The study also offers a technical classification of radar types and configurations (e.g., monostatic vs. bistatic), focusing primarily on system functionalities. The issue of falls in elderly individuals is also examined in [7], where a Frequency Modulated Continuous Wave (FMCW) radar is applied in AAL and clinical practice for long-term balance monitoring. Although conducted in a controlled laboratory setting with predominantly young participants, the system achieved a 98.4% accuracy rate in identifying foot movements during balance tests. The paper discusses ethical and privacy concerns associated with the real-world deployment of radar systems, stressing the importance of developing guidelines to protect individual privacy. However, it does not explicitly conclude that radar is privacy-friendly. The work in [8] investigates mmWave radar for monitoring patient safety in high-risk mental health settings, showcasing its application in clinical environments. The paper emphasizes the contactless nature of mmWave radar, which enhances privacy by eliminating the need for users to wear or carry devices. Another recent study, ref. [9], focuses on elderly care and AAL in low and middle income countries. It employs FMCW radar sensors integrated with deep learning techniques for non-contact and non-invasive monitoring, applicable to telemedicine, telehealth, and teleassistance. The work positions radar as an ethical solution for healthcare and public safety. However, unlike our manuscript, these studies do not provide a comprehensive ethical analysis or explicitly address Key Value Indicator (KVI) such as fairness, inclusiveness, sustainability, or implications for long-term care. Their ethical discussions are largely limited to privacy and are often framed in terms of comparing radar technologies to other solutions with regard to accuracy and intrusiveness. In contrast, our manuscript introduces the novel concept of KVIs as a framework to align technical development with broader societal goals. These examples, not only from recent scientific literature, are summarized in

Table A2. Comparative table of recent studies.

Ref.	Summary	Limitation	Ethical Issue	AAL	KVI
[52]	mmWave radar aids health monitoring and activity recognition. A 4D TDM MIMO radar model was established using the HDM05 dataset. Simulator achieves 90% classification accuracy with PointNet. Compares pre-processing and point cloud methods for activity classification. Focuses on chirp signal for MIMO radar applications. TECHNOLOGY: 4D TDM MIMO radar.	Features must be handpicked, limiting algorithm universality. Small sample size may affect training accuracy. Non-uniform training and test set distributions can lead to errors.	The paper discusses privacy in relation to radar technology, highlighting its non-intrusive sensing capabilities, so it is considered privacy-friendly	YES	NO
[12]	The study uses deep neural networks to classify human activities from radar point clouds, achieving high accuracy with 2D-DCNN and LSTM. Limitations include distance constraints and limited data diversity, highlighting the need for further research TECHNOLOGY: MIMO FMCW radar (77 Ghz)	Limited angular resolution affects target recognition at a distance. - Only 19 subjects were measured. - Micro-Doppler signatures are ineffective for random human motion.	Not addressed.	NO	NO
[26]	The paper proposes mmWave MIMO radar for detecting moving people. It processes range and data cube information simultaneously for improved detection. Experimental results validate the effectiveness of the proposed method in various scenarios. TECHNOLOGY: mmWave MIMO radar.	YOLO struggles with range-azimuth maps, leading to false negatives. Common evaluation metrics like classification matrices are unsuitable for this method. Traditional CFAR methods require careful parameter selection, complicating detection.	The paper discusses privacy concerns related to sensor technology. The use of radar technology is highlighted as beneficial in dark conditions, but the problem of privacy remains unresolved.	NO	NO
[14]	The paper presents the EthicNet architecture, which integrates (KVIs) with (KPIs) in network services. It leverages User Digital Twins to personalize KVIs, addresses stakeholders’ ethical requirements, and identifies open challenges in ethics-aware networking.	Conflicts may arise between environmental sustainability and profit requirements of service providers. New metrics are needed to evaluate performance against key value indicators. Resource sharing must adhere to economic principles to reduce energy consumption.	The paper highlights key ethical issues in network services—such as sustainability, trust, and fairness—and suggests using KVIs with performance metrics to ensure alignment with user values and well-being.	NO	YES
[3]	The paper proposes a KVI framework for ICT R&D value analysis. It emphasizes aligning technology development with societal values and sustainable goals. The framework includes five steps for assessing value outcomes.	Conflicting values between stakeholders complicate value identification. Organizational culture may hinder designers’ ethical responsibility. Flexibility in value interpretations can lead to ethics-washing.	The paper discusses the importance of integrating values into ICT development to enhance usability and privacy, while also addressing ethical responsibilities. It emphasizes the need for a values-based approach to address societal challenges.	NO	YES
[53]	AAL technologies support elderly quality of life and reduce caregiver burdens. Ethical issues arise from AAL technology use in personal spaces. The paper advocates for a person-oriented approach in healthcare design. TECHNOLOGY: AAL technology in general.	Age-related shortcomings require innovative processes for the elderly. Ethical bias may distort AAL tech effectiveness. Population bias affects external validity of findings. Unintended consequences may arise from interventions. Ethical principles face challenges in practical application.	The ethical issues identified include health and well-being, trustworthiness, usability, privacy concerns, accuracy, fairness, economic and environmental sustainability. While these are a significant concern in AAL, the specific inquiry regarding radar technology is not addressed.	YES	NO
[43]	The paper explores visual privacy protection methods in assisted living technologies. It emphasizes the importance of Privacy by Design in safeguarding data. Researchers advocate for understanding user needs and the long-term effects of technology. TECHNOLOGY: video-based monitoring systems (VMS).	The qualitative methodology limits generalizability to other environments. Short-term system usage restricts comprehensive understanding. Limited assistive technology literacy among older adults may create resistance.	The paper emphasizes the importance of engaging users to ensure that solutions are ethically sound and address genuine needs, which relates to economic and environmental sustainability. There is no mention of radar technology or its privacy implications, as this paper focuses on visual-based monitoring systems.	YES	NO
[48]	The paper explores bias and discrimination in AI from multiple perspectives. It highlights challenges in creating nondiscriminatory algorithms. Digital discrimination affects areas like credit and policing. Current laws struggle to address algorithmic discrimination. TECHNOLOGY: AI algorithms.	Valid causal graphs for bias assessment are often unfeasible. Current legislation struggles with algorithmic discrimination. Access to training data is frequently restricted. Digital discrimination remains a challenge for technical and ethical reasons.	The ethical issues related to AI include the need to reassess moral standards due to discriminatory risks. These impact fairness and trust in decisions affecting health, well-being, and care. Economic and environmental sustainability are also considered.	NO	NO
[54]	The paper discusses societal challenges in 6G development. It emphasizes a value-based approach alongside performance-driven technology. KVIs are defined for monitoring societal impacts. The UN SDGs framework is utilized for assessing societal values. 6G aims to address societal needs and create value. TECHNOLOGY: 6G		The research paper emphasizes the importance of societal values, including health and well-being, in 6G development. It argues that these values should guide technology to ensure positive societal impact. It also discusses economic and environmental sustainability through KVI monitoring.	NO	YES
[55]	6G vision focuses on new business models and human possibilities. Key technologies include THz spectrum, AI, and edge-centric architecture. Performance indicators target latency and reliability. 6G aims for digital inclusion and sustainability. Automation and network slicing will decouple costs from growth. TECHNOLOGY: 6G	The paper addresses privacy, particularly in the context of 6G systems and sensor design, which must respect user privacy.	No specific mention of radar technology is provided.	NO	YES
[56]	The document analyzes obstacles to 6G feasibility. It discusses societal expectations and feedback on advanced communication services. The paper outlines business models for the remaining representative use cases. It proposes strategies to enhance the sustainability benefits of 6G solutions. TECHNOLOGY: 6G	Resource-efficient communication systems face significant development challenges. Recycling processes struggle with technological evolution and device complexity. Spectrum management limitations exist across regions. Large investments in infrastructure may deter stakeholder profitability.	The paper addresses health and well-being risks from technology use, including potential harm and disinformation. It highlights trust, data privacy, and stresses economic and environmental sustainability through responsible practices.	NO	NO
[16]	6G combines positioning and sensing to enhance communication, with a focus on sustainability, inclusiveness, and trust. It supports global access, uses cooperative networks to reduce costs, and links societal values to KPIs through KVIs. TECHNOLOGY: 6G	AI mechanisms need scrutiny due to opacity. Vulnerabilities exist in GNSS and UWB. Every measurement and hardware in 6G has potential security weaknesses.	The paper discusses the ethical implications of positioning and sensing in 6G, focusing on quality of life, especially in elderly care and patient supervision. It covers health, privacy, accuracy, trust, and sustainability.	NO	YES
[8]	The paper proposes a low-data solution for mental health settings. It utilizes a Kalman filter to improve target tracking accuracy. Results show reduced false positives and negatives in human tracking. TECHNOLOGY: mmWave radar.	The sensor provided only target coordinates, not raw data. Tests with a single participant limit generalizability. The model misclassified slow targets and was limited to 2D tracking without consistent target count handling.	The paper emphasizes privacy in patient monitoring, noting that non-visual sensors protect dignity. It highlights the importance of accuracy for trust and mentions the economic sustainability of mmWave radar.	YES	NO
[9]	The paper explores radar technology for non-invasive elderly monitoring. It highlights ethical considerations in ambient assisted living. A deep learning approach enhances radar gait analysis. Challenges of low and high frequencies are discussed.	Radar can be difficult to interpret due to background noise. Noise removal thresholds are arbitrary. Monitoring may inhibit natural behaviors.	The paper discusses privacy in telecare, highlighting the importance of fairness, integration, and sustainability. Radar is seen as privacy-friendly as it does not reveal identity traits.	YES	NO
[7]	Addresses fall risk monitoring with FMCW radar in ambient assisted living and clinical practice. Emphasizes low-cost, long-term balance assessments. Machine learning integration improves accuracy. TECHNOLOGY: FMCW radar.	Conducted in a lab setting, limiting real-world use. Young participants limit generalizability. Distances beyond 1.25 m impact accuracy. Misclassifications in movement predictions.	Emphasizes privacy, trust, and accuracy in radar-based health monitoring. Stresses need for ethical deployment guidelines.	YES	NO
[28]	Explores radar-based activity recognition using AI. Highlights innovative validation methods. Experiments include machine learning for radar data. A comprehensive literature review is included. TECHNOLOGY: radar + AI.	Model not tested under occlusion. Sparse radar data limits classification. Dataset is small. Evaluation only on limited simulated space.	Paper emphasizes ethical considerations in Human Activity Recognition, especially data privacy and participant rights. Stresses growing importance of ethics as radar spreads.	YES	NO

, in Appendix A to highlight thematic overlaps and key limitations, especially those compared to the topic addressed in this work. Recently, a review evaluated typical machine learning models for the analysis and the extraction of vital signs, focusing on the technical issues related to the datasets, metrics and methodologies [10]. The data scarcity problem in the field of millimeter wave sensing is an issue limiting the proper expansion of the field when compared to other sensing technologies such as RGB cameras. There are attempts to address the gap by adding data to fuel the research, and the reader can find a deep investigation of this issue in [10], where publicly available datasets are analysed in terms of size, gender and age distribution. Typically, the sample size of the publicly available dataset is around 30 subjects with balanced gender diversity. The age distribution is biased towards young adults, and there is still no dataset regarding the elderly population [10]. Remote monitoring with radars can be seen in a broad set of applications, encompassing classification of activity of daily living, fall detection, classification and measurement of vital signs, gait analysis, and indoor localization and tracking, demonstrating great versatility and sensitivity to micro and macro motions of a subject. In refs. [11,12], the value of accuracy for the classification of human activities with a micro-Doppler is, respectively, 95.3% and 97%. Typically, the engineering papers developing approaches in the aforementioned aspects are focusing on the advancement of performance metrics, showing a methodology that can reliably work under certain circumstances in a vicious circle of continuous improvement from previous state-of-the-art methods. This interest in the performances of the ICT derives from the formulation of the Key Performance Indicator (KPI) of a technology, which are clear, quantifiable, verifiable, and measurable objectives, bounded by a policy of continuous improvements. As defined by the ISO standard 9001:2015 [13], KPI should be monitored, communicated and updated. The issue with these criteria is that they should be tailored for the specific application and the specific technology, and given the policy of continuous improvements, they are only short-term goals that cannot reflect the relevance of the technology and the possible societal benefits that the technology is achieving long term.

This table systematically maps the core KVIs addressed by FMCW radar in Ambient Assisted Living, such as productive aging, community-based care, user-centered systemization (usability, privacy, fairness, accuracy), and sustainability innovation, which will be explained in the next section and shown in Figure 1 with relevant clauses within the ISO 9001:2015 quality management framework. By doing so, we align our value-driven technology assessment with internationally recognized quality control principles, supporting future standardization and deployment efforts. The standard core is the implementation flow PDCA (Plan-Do-Check-Act), that consists of:

• Plan: Establish requirements: privacy, usability, detection, sustainability,
• Do: Implement radar modules, micro-Doppler analytics, integration,
• Check: Monitor performance indicators; user studies, field data,
• Act: Correct nonconformities; update design; scale successful modules.

The ISO procedures mentioned in

Table 1. Correlation between FMCW Radar KVIs and ISO 9001:2015 procedures.

KVI Theme	Radar KVI Examples	ISO 9001:2015 Clause	Implementation via ISO Procedure
Productive Aging and Community-Based Services	Fall detection accuracy, gait analysis, vital signs monitoring	88.1, 8.2 (Operational planning and service requirements); 9.1 (Performance evaluation)	Service design and planning; measurement of outcomes and satisfaction
Usability, Privacy, Fairness, Accuracy	Non-imaging sensing, low false positives, high detection precision, novel micro-Doppler analytics	4.4 (System approach/PDCA), 5.1/5.2 (Leadership commitment to policy) 9.1.3; 10.2 (Nonconformity and corrective); 6.2/6.3 (Objectives and planning); 10.3 (Continual improvement)	Leadership commitment; defined objectives (accuracy, privacy); root cause and corrective action processes, Setting innovation goals; system-based management; continuous improvement through PDCA, harvest data for improvement
Economic and Environmental Sustainability	Cost-effective components, low energy use, modular scalability, low-power designs, eco-sourcing	6.1 (Risk and opportunity), 7.1.3 (Infrastructure and environment), 10.3(Improvement for suitability/adequacy)	Risk and opportunity analysis/management; efficient infrastructure; sustainable development planning

are:

Clause 5—Leadership and Policy: Top management must define a quality policy that supports accuracy, privacy, innovation, sustainability, and communicate it across functions.

Clause 6—Planning (6.1–6.3):

• 6.2: Align measurable objectives (e.g., ≤5% false alarms, ≥90% fall detection).
• 6.1: Evaluate risks/benefits of new eco-materials or energy-saving circuits.
• 6.3: Plan changes to integrate community-based tech or test new algorithms.

Clause 8—Operation

• 8.1/8.2: Define functional design, usability acceptance, and field validation.
• Releases: Implement checkpoint-based functional tests and usability reviews.

Clause 9—Performance Evaluation:

• 9.1.3: Analyze data from deployment: false alarms, privacy concerns, detection accuracy.
• 9.3: Include KVI results in management review to inform decisions and resource allocation.

Clause 10—Improvement:

• 10.2: Root-cause investigations of misdetections or privacy complaints.
• 10.3: Leverage PDCA for iterative improvement of algorithms, power, and materials.

By doing so, we align our value-driven technology assessment with internationally recognized quality control principles, supporting future standardization and deployment efforts.

A clear recent trend that can be observed in rapidly growing sectors, such as in the case of the telecommunication field, is to consider new value-based long-term goals as the basis of the evaluation of technological solutions [14]. For example, the sixth generation of mobile communications is currently being developed with novel goals and challenges in mind, such as Sustainability, Inclusion and Reliability [15]. This aspect is also addressed in [16], where the authors analyze emerging networking technologies “through the lens of sustainability, inclusivity, and trustworthiness”. These technologies aim to leverage communication networks not only for data transmission but also for simultaneous environmental sensing (Integrated Communication and Sensing, ISAC), a topic widely discussed in current scientific literature and expected to be concretely implemented with the introduction of 6G networks. The study emphasizes the relationship between KVI and KPI, proposing a framework in which the value indicators are quantified based on the outcomes driven by performance indicators. This approach ensures that the objectives of these technologies are aligned with the principles and goals of the United Nations’ Sustainable Development Goals. The new paradigm for the design, development, and validation of ICT technology is shifting from performance to the consideration of adding value to society and promoting well-being. The question is not only whether or not a system is working properly and under certain performance constraints but also whether or not society can benefit from the technology and to what degree.

As a consequence, in a pyramidal organization of goals, the peak would be society’s well-being and the KVI are the objectives needed to achieve it. The KPI are instead the foundational low-level objectives that enable the proper functioning of technology and the achievement of high-level value objectives. Even before defining specific technological milestones, there should be a clear understanding of the qualitative values that the technology aims to achieve.

Physical and cognitive impairments are typically associated with the aging process, leading to a status of frailty, with a loss of autonomy and independence; as a consequence, elders need constant help from specialized healthcare staff and eventually experience a decline in social participation and quality of life. These issues fuel the research on novel assistive technologies that could support and improve the engagement of elder individuals in their normal routines at home, maintaining their sense of freedom and independence. Assistive technologies should focus on the capabilities, rights, and desires of the elderly, as well as on issues related to safety, protection, harm prevention, risk reduction, and chronic disability management. By increasing autonomy, self-confidence, and mobility, and promoting active lifestyles, assistive technologies are expected to reduce the risk of disability and institutionalization, improve safety, prevent social isolation, and maintain support networks, thus enabling the elderly to age in place.

Different generations of assistive devices have provided support to individuals in need, achieving improvements in accessibility, usability, acceptability, and trust each time. The current state-of-the-art technology provides ambient sensors that can be seemingly integrated with the environment, achieving what can be called an AAL [1,17] and ensuring a normal autonomous and active life into old age, improving the quality of daily living, and reducing the burden of healthcare and social care.

2. Comparisons of Radars with Other Technologies

The possible technologies that can be used in AAL can be distinguished in contact and non-contact devices [5]. Contact devices include wearables, such as smartwatches, and biomedical sensors in the form of cardiorespiratory belts and electrodes, whereas non-contact devices include RGB cameras, RGBD cameras, radio-frequency (RF) sensing and radars, acoustic sensors, infrared cameras, and thermal cameras. This first distinction is relevant because to achieve a proper intelligent ambient, where a subject can be monitored non-intrusively and continuously in daily activities, one or more ambient sensors are necessary. FMCW, with mmWave bands, offer very short wavelengths based on central frequency (5 mm and 3.9 mm, respectively), which provide high spatial resolution and strong sensitivity to small movements. Higher frequencies allow the system to detect defined spatial details, such as subtle chest displacements during breathing. Moreover, because of the short wavelength, even movements—like those caused by respiration, heartbeat, or postural sway—produce measurable phase shifts in the radar signal. These properties make mmWave radars particularly effective for accurate and contactless monitoring of physiological and behavioral parameters. Moreover, wearables often face the reluctance of the subject due to limited comfort in wearing them or sensitive skin and need the subject’s cooperation. A comparison of sensors is reported

Table 2. Comparison table of various sensors in terms of centre frequency, sensing mode, and maximum range.

	Central Frequency	Max Range	Depth Sensing Technology	Advantages	Disadvantages
Radar FMCW	Ghz	150 m	Frequency modulation and signal processing	robustness to environmental changes, direct evaluation of velocity with Doppler shift	low angular resolution and small interpretability of the maps
Lidar	kHz	50-100 m	Time of flight sensor	high angular resolutions, point cloud visualisation	no information related to velocity
Depth Camera	Thz	10 m	Time of flight, light patterns or stereoscopy	high angular resolution, point cloud visualisation	limited range, no information related to velocity

.

Figure 2 shows a radar plot proposed in a recent review [5] on the typical AAL technologies and how they can be scored based on privacy level, data processing, and integration in the home tech.

It is possible to observe that there is already a clear interest in the topic of privacy for the application of assistive technology, even though this is still not a completely investigated theme. From the figure, it is possible to notice that radars and RF sensing are considered privacy-friendly technologies, but they are still not well integrated into home technologies, motivating the research line of radars applied to AAL environments.

The use of radars in ambient assisted living environments has already been explored in different review papers. In this paper, the focus is on radar sensors rather than wearable ones, as they allow for unobtrusive monitoring without requiring the person to wear or interact with any device, something that could cause discomfort or lack of compliance. Moreover, radar sensors do not provide direct information about personal characteristics unless specific algorithms are applied, which helps preserve privacy, unlike monitoring done with RGB sensors. A practical motivation is that the radars used in our work are for the automotive area, making them easily available on the market and relatively low-cost. Shah and Fioranelli presented a set of challenges of RF sensing in AAL where it is possible to observe the topic of robustness, reliability, affordability, and ethical concerns, such as privacy and exposure to electromagnetic radiation [18].

Nevertheless, the focus of the review was on the technological aspect; therefore, typical data processing was explored, a categorization of radar technologies was proposed, and an analysis of trends in the topics was performed. Recently, a review explored the advancements in signal processing for the extraction of human motion from radar point clouds for applications in the AAL environments, such as activity and gait recognition, fall detection, and posture estimation [19]. This interest in radar signal processing for healthcare-related applications is evident from the previously cited works; however, the point of view of the KVI has seen limited attention.

This paper aims to introduce a novel paradigm in the form of the key values that should be considered whenever dealing with a technology applied to elderly assistance. In developing new assistive technology, short-term and long-term goals are often needed to help define what the technology should achieve. Short-term goals in terms of performance metrics and advancements in signal processing have been exhaustively explored in other reviews, but the possible long-term goals that justify the effort for continuous improvement are under-researched.

The long-term goals that radar should accomplish are first examined in Section 3, followed by a tutorial on how current radar systems work in Section 4 and what values could be achieved thanks to the adoption of the millimeter-wave sensing in Section 5.

3. The Choice of Key Values for the Aging Society

In September 2015, 193 leaders met at the United Nations (UN) to approve the 2030 Agenda, an ambitious program to be completed within the next 15 years, consisting of 17 sustainable development goals characterized by universal validity, requiring all participating countries to contribute to their achievement according to their capabilities [20,21,22]. These objectives aim to end poverty and inequalities and promote social and economic development, in peaceful contexts, without neglecting environmental sustainability. Among the defined goals in Agenda 2030, goal 3, which is “Ensure healthy lives and promote well-being for all at all ages”, is the one mostly related to elderly care and AAL, and gives relevance to the issue of a healthy and active aging society. Additionally, the ten years before 2030 have been declared by the United Nations as the decade of healthy aging [23] to gather interest in the issues society needs to face in the near future.

A starting point to understand the challenges that our society is facing and will face if the issue of the aging population is not properly and promptly addressed are discussed in ISO IWA 18:2016 [24], providing possible guidelines for the development of solutions. In particular, the document analyzes aspects related to health, social care, and well-being by establishing five principles that the services provided to the elderly should respect and promote:

• Human Dignity: The first principle of human dignity can be achieved with an improvement of social participation, as it gives a role in society to the elder subject.
• Productive Aging: This second principle is directly connected to the quality of life into old age and as a consequence to an improvement of health and well-being;
• Community-Based Service: This third principle regards the improvement of assistance and care through the help of community initiatives.
• Human-Centric Design: The fourth principle adds the necessity of a human-centric design in the proposed solutions. The service should not undermine the independence of the elderly subject, who should be able to live normally. To achieve this, the service should be trusted and fair.
• Innovation for Sustainability: The fifth principle is connected to the principle of continuous improvement. This is also the principle that is tightly related to technological improvement due to the pursuit of innovation while taking into consideration all the possible consequences for the economic and environmental sustainability.

The mentioned principles are generally defined for any service provided for elder care; however, they are going to be applied in pillars to be achieved through millimeter-wave radars, as shown in the Table 1.

Our approach begins by identifying a subset of universally shared values from the United Nations Sustainable Development Goals (SDGs), which represent ethical and social priorities across all sectors. Among these, we focus in particular on Goal 3—Ensure healthy lives and promote well-being for all at all ages, as it aligns closely with the objectives of AAL technologies and with our specific context of use: supporting elderly and frail individuals. For supporting the elderly, continuous monitoring enables the collection of information such as movement speed, respiration rate, and other vital parameters. This allows the early detection of potential issues, such as falls, thereby helping to prevent hospitalizations. Fall-related problems are particularly critical in the elderly population, making such monitoring especially valuable. From this starting point, we progressively translated these universal principles into concrete, everyday dimensions such as autonomy, safety, non-invasiveness, and privacy, with the aim of bridging high-level ethical goals with the actual functionality of radar systems. We define KVIs as system-level parameters that operationally reflect those values and can guide both the design and the evaluation of technological solutions. In the present review, however, these KVIs have been identified in a qualitative way, as functional properties and contextual principles that express the alignment between FMCW radar systems and relevant ethical-social values. We acknowledge that a major future challenge will be to define and formalize quantitative metrics for KVIs, based on shared rules and mathematical models, so that they may be considered directly in the design phase. This will be essential to enable more value-aware, measurable, and responsible technology development.

4. An Overview of Radar Working Principle and Radar Signal Processing

This review focuses specifically on FMCW radar systems due to their prevalence in both academic and commercial ambient assisted living applications. FMCW radars represent the most commercially mature and technically balanced solution for indoor monitoring, offering the ability to simultaneously estimate target range, velocity, and angle using compact, low-power hardware.

Alternative radar types, such as Continuous Wave (CW) and Utra Wide Band (UWB), have been explored in specific research contexts but present limitations for general AAL deployment. Continuos Wave (CW) radars lack range resolution and are thus unsuitable for spatial tracking, while UWB radars are generally more complex to implement.

FMCW radars have gained widespread adoption in recent years, notably increasing their use in applications related to AAL [10]. These systems operate by emitting a continuous sinusoidal signal, known as a “chirp”, whose frequency linearly increases over time. This frequency modulation distinguishes FMCW radars from traditional cw devices and provides the capability to simultaneously measure the distance and velocity of targets. FMCW radars contain the following components, including a synthesizer, a transmitting antenna, a receiving antenna, and a mixer. The synthesizer generates a chirp signal as shown in Figure 3, which is transmitted by the transmitter (TX) antenna and hits an object, the reflection of the chirp is detected by the receiving (RX) antenna; the received signal is mixed with the transmitted signal to obtain an output sinusoidal signal with an “intermediate frequency”, calculated by the difference between the frequencies of the two input sinusoids and a phase, given by the difference between the phases of the two input signals. Since the reflected signal has traveled to the object and back, there will be a delay between the transmitted and received signals. The distance to the object can be determined by measuring the difference in frequency between the transmitted and received signals. This is because the time delay causes a frequency shift that is proportional to the distance. Detecting velocity is necessary to transmit a sequence of chirps, forming a frame of the acquisitions. The chirps are divided by an inter-chirp time duration, which can be set to tune the maximum measurable velocity. The total number of the chirps inside a frame is always inside a time interval of choice, called periodicity, as can be seen in Figure 4, which corresponds to the time duration of one frame.

Thanks to the frequency modulation of the signal and to the Multiple Input Multiple Output (MIMO) technology inside, the FMCW radar can detect three fundamental parameters: range, Doppler, and azimuth. The range represents the radius that measures the distance of an object or subject from the radar. The Doppler shift is related to the velocity of the object or subject. The azimuth angle is related to the azimuthal angle of arrival of the reflection and gives information on the horizontal position of the object or subject.

Data collected by FMCW radars are typically organized into three-dimensional matrices, often visualized as data cubes. These cubes are generated by applying a two-dimensional Fast Fouerier Trasform (FFT) across different axes of the raw data. Each face of the cube represents a distinct projection, forming what are known as radar maps: the range–velocity (or range–Doppler), range–angle, and angle–velocity (or angle–Doppler) maps. In many cases, the angular component refers predominantly to azimuthal angles, primarily because this information is crucial for reconstructing a spatial, Cartesian representation of a detected object’s position—essential for localization and tracking tasks [26].

Converting the polar coordinate-based maps into Cartesian coordinates yields what is termed the XY-range map, as shown in Figure 5. This conversion enables understanding of the radar’s field of view in Cartesian coordinates, which is a more interpretable way to display the subject position rather than polar coordinates.

When temporal analysis is incorporated, these two-dimensional radar images provide insights into how the environment and targets evolve over time. This dynamic data facilitates a wide range of applications, including activity recognition, precise localization, motion tracking, and vital sign detection.

Within these applications, two key Doppler-related quantities are of particular importance: micro-Doppler and macro-Doppler effects. The Doppler phenomenon, directly related to target velocity, offers valuable clues about movement patterns. Micro-Doppler effects capture subtle oscillations caused by fine movements—such as limb motions or breathing, while macro-Doppler shifts are associated with broader, more substantial displacements, such as overall body motion. These maps portray the entire monitoring zone, determined by the radar’s viewing angle in conjunction with its maximum detection range. The reported scenario involves a subject walking within a cluttered environment, demonstrating the radar’s capacity for tracking movement amidst obstacles and complex surroundings

Velocity has been previously identified as a measurable parameter and is visually represented through two supplementary maps: the X-range velocity map, as shown in Figure 6, the Y-range velocity map, as shown in Figure 7.

These X and Y velocity maps simultaneously show an object’s coordinates in the radar Cartesian reference system and its relative speed. Both maps feature a zero Doppler zone, which indicates the presence of objects with no relative motion either approaching or receding from the radar. This zero Doppler is a threshold separating positive and negative velocities: positive velocity values correspond to objects moving toward the radar, whereas negative velocity values reflect objects moving away from the radar.

In Figure 8, analyzing temporal information, it is possible to extract from the Doppler-range map and the angle–Doppler map information about the variation in the speed of a person’s walk for gait analysis evaluation or possible fault detection. These visualizations are obtained by displaying the evolution of a single parameter over the time of an entire acquisition. Starting from the radar data cube, it is possible to analyze the temporal evolution of three extracted features: velocity, azimuth, and range. To obtain this type of representation, one can employ a spectrogram-based approach. For example, by considering all the range-Doppler maps acquired over time, each map can be compressed into a one-dimensional vector by aggregating along a selected dimension.This process results in a sequence of vectors, one for each frame, forming a matrix where one axis represents time and the other corresponds to the other dimension. Applying a spectrogram (e.g., via short-time Fourier transform) to this matrix allows for visualizing the temporal evolution of the signal content, highlighting changes in velocity, range, or angle over time.

Commercial Radar Sensors

The increasing availability of millimeter wave radar modules on the market, mainly driven by their widespread use in the automotive sector, has been an enabling factor for the implementation of radar-based solutions also in the field of ambient assisted living. Automotive applications, such as adaptive cruise control and braking assistance systems, have driven the development of compact, high-performance, and cost-effective radar sensors. This has led to the standardization and large-scale commercialization of FMCW radars, making them accessible for indoor use as well. This availability has facilitated technology transfer to the home care and healthcare sectors, resulting in the emergence of several radar devices specifically designed for AAL. One example is the Milesight VS373, a 24 GHz radar sensor with LoRaWAN connectivity, capable of detecting falls and presence in residential environments without the need for wearable devices or cameras. The VS373 features an MIMO antenna array with 24 transmitters and 22 receivers, operating at 24 GHz with a detection field of 70° horizontally and 140° vertically. It achieves fall detection accuracy up to 99% with a false alarm rate below 5%, supports LoRaWAN protocols for long-range wireless communication, and has an IP65 rating for dust and water resistance. Another example is the XGZP6867- Fall Radar module, available through consumer platforms like AliExpress, reflecting the rise of low-cost radar solutions dedicated to home security and individual autonomy monitoring. This compact module operates at 24 GHz using FMCW modulation and communicates via an UART interface. It is powered by 5V DC and is suitable for integration into portable or wearable devices for posture monitoring and fall detection in domestic environments. Finally, the multinational Hikvision has developed a line of AI-assisted radar systems aimed at continuous monitoring of vulnerable users. These devices combine behavioral analysis algorithms with mmWave sensors employing FMCW modulation and digital beamforming. They can detect falls, immobility, and movement patterns while also monitoring vital signs such as respiratory and heart rates. Designed with privacy in mind, these systems do not record images, ensuring user confidentiality. The Hikvision solutions are compatible with centralized management platforms like HikCentral Professional, allowing integration in nursing homes, hospitals, and home care settings for reliable, continuous monitoring. The adoption of these devices highlights the transition of mmWave radars from embedded automotive components to enabling technology for person-sensitive applications, with direct implications in terms of accessibility, scalability, and operational reliability in AAL contexts.

5. Radar Key Values and Relative Key Performances

In the following sub-sections, the previously established five principles for AAL services are explored in the context of the application of an FMCW radar system with the support of state-of-the-art scientific literature.

5.1. Productive Aging: Health and Well-Being

The first principle of productive aging can be achieved with healthy aging and an improved quality of life into old age. Physiological signal sensing is the primary method for tracking a subject’s health and well-being. As evidenced by recent studies of the scientific literature [10], radars have found increased use in the field of vital sign extraction due to the development of machine learning models. Radars are sensitive to velocity and distance, as was previously indicated. The velocity information is associated with a shift in the phase of the received chirp signal, and this shift is sensitive to tiny, recurring displacements characteristic of the pseudo-periodic chest movement brought on by breathing and heartbeat. The field of vital sign sensing could be divided into the estimation of point values [27], such as heart rate, respiration rate, blood pressure, and arterial oxygenation [28], and reconstruction of an interpretable signal [29], such as a breathing signal or a cardiac waveform, and the detection of physiological events [30], such as R-peaks, or heart sounds or pulse waveforms of an electrocardiogram. The ability to continually monitor these vital signs, together with the radar recordings’ non-intrusive nature, can help families and caregivers identify and address cardio-respiratory issues before they develop. Furthermore, in this article [31], it is demonstrated that it is possible to compare contactless sensors (like radar or RGB sensors) with a Polar used as a reference; Radar-based physiological signal monitoring can also be used to identify a user’s emotional state, according to another novel line of study [10]. Commonly, affective computing is based on either the classification of a discrete set of basic emotions, such as anger, surprise, fear, or disgust, or on the regression of continuous dimensions that compose an emotion, such as arousal, valence, and dominance. Both approaches have been used in the scientific literature. Recognizing the emotional status of a frail subject is relevant for adequate human–machine interactions and is necessary to deal with the high prevalence of major depressive disorder among elders [32]. Therefore, a mental health assessment needs to be included in the regular health screening program. However, compared to other areas, affective computing in radar sensing is still not well studied [10].

In the context of AAL, the technology has utility since it allows for the constant and non-intrusive assessment of the subject’s emotional state. Radars are devices that can assess the external behaviors of the subjects, which could be related to certain emotions [33] and can also detect slight oscillations of the torso due to heartbeat and respiration to recover vital signs features related to different emotions [34]. In this way, radars could be viewed as instruments that can monitor remotely both the external and internal loads of individuals, through the assessment of macro-motions and micro-motions, respectively.

In Figure 8, it is possible to observe how the previously explained radar working principle (Section 4) can be employed for the evaluation of macro and micro motions for the evaluation of the functional status or the cardio-respiratory status or a subject, respectively. The applications of fall detection and gait analysis using radar are based on the temporal analysis of the data collected by the radar. The following is an explanation of how they work. Regarding fall detection, the radar measures movement and detects changes in body position. The data is analyzed in the time domain to identify rapid, abnormal changes in body movement, which are typical of a fall. Regarding gait analysis, the radar collects information on the movement of the legs, arms and body during walking; in this way, parameters such as speed, step length, and step frequency could be extracted. Finally, for the vibrational analysis, a specific bin within the azimuth range is selected (the bin is the same for the entire acquisition), and from this, coming back to the raw radar data cube, a temporal vector representing the phase variation in the radar signal is extracted. These changes correspond to chest movements caused by breathing and heartbeat. Analyzing the final results, it is evident that radar images respect privacy regarding the specific conditions of a person. In fact, it is not possible to determine how the monitored person is dressed or their skin color, facial expressions, or body shape. This is an intrinsic advantage of radar technology, as such privacy preservation is achieved without the need for additional mitigation algorithms. Of course, it is important to note that radar data can be used to recognize individuals through their micro-Doppler signatures. However, personal attributes remain anonymous, and in a monitoring context, it is more relevant to detect and evaluate potential issues that can be inferred from the micro-Doppler patterns. Understanding the technical mechanisms behind radar-based applications allows for a more informed analysis of their impact on the Key Value Indicators (KVI), thus providing an integrated perspective between technological performance and societal value. By analyzing the frequency and intensity of these vibrations, it is possible to gather information on the respiratory and heart rhythms.

• Fall detection and gait analysis enhance health and well-being, promote productive aging, and support community-based services by enabling early detection of health deterioration and reducing emergency response times.
• Vital sign monitoring supports non-invasive care and trustworthy systems (through continuous, privacy-respecting observation), while also aligning with economic and environmental sustainability by eliminating the need for wearable devices or high-maintenance systems.

These direct links between radar capabilities and societal goals reinforce the systemic value of radar-based AAL technologies.

5.2. Community-Based Services: Assistance and Care

The second principle of community-based services can be reached by improving the field of elderly assistance and care by integrating ambient sensors to monitor a subject continuously and non-intrusively. Assistance and caregiving tasks encompass the monitoring of elderly individuals’ daily activities, particularly those activities that require support and cannot be performed independently. Furthermore, providing precise, real-time information regarding the subject’s position, movements, and activities can facilitate caregivers and families in better managing their responsibilities and alleviating their workload.

The integration of millimeter-wave radar within a domestic environment offers the potential to deliver healthcare capabilities directly to the families, such as tracking the motion of the individual [26], monitoring activities of daily living [35], fall detection [36], and gait pattern analysis [37].

The localization of users indoors through sensor-based approaches remains an ongoing challenge; however, radar sensors have demonstrated effective capabilities to accurately detect the position of a subject within indoor spaces [26]. Human activity recognition, a prominent area of research in mmWave technology, aims to identify and verify the performance of routine daily activities [35]. Ensuring that elderly individuals remain active and engage in typical daily behaviors is critical for early detection of frailty and related health decline [38].

Radar sensors generally classify various activity types by generating radar maps that show the evolution of the subject’s velocity and limb movements over time (velocity-time maps) and spatially (range-velocity maps), as shown in Figure 9.

Fall detection is closely linked to activity recognition, as falls can be identified through similar data acquisition processes and algorithmic frameworks. Consequently, it is feasible to develop radar-based systems capable of continuous monitoring and alerting caregivers or family members in the event of a fall occurrence. Alternatively, these systems can synthesize routine activity patterns to assess potential declines in autonomy or cognitive functions by analyzing gait parameters and spatiotemporal gait features, such as gait speed [37].

5.3. Human-Centric Design: Trustworthiness

Trust in ICT is a complex and multidimensional concept that goes beyond the typical security measures. According to ISO/IEC 5723:2022, the trustworthiness of a system is defined as “the ability to meet stakeholders expectations in a verifiable manner” [39]. Different aspects come together to achieve a trustworthy, and across all the possible qualities, we decided to explore the field of usability, privacy, and accuracy of radar systems and how these concepts could provide a way to achieve greater trust in the application of the technology.

5.3.1. Usability

Usability is a dimension of trust that can be explored when dealing with remote sensors. Given the potential mobility and access challenges faced by elderly individuals, it is crucial to develop technologies that are mobile, manageable, and easily replaceable. These solutions should be non-wearable and adaptable to their changing needs and health conditions. Usability, as outlined by [39], refers to the degree to which a technology enables users to accomplish a specific goal efficiently, effectively, and satisfactorily. In this context, radar systems are recognized for their versatility as contactless sensors, capable of being installed in various structural environments without requiring significant modifications. Typical room dimensions in residential settings—approximately 10 (m²—are well within the detection performance range of radar systems commonly explored in scientific studies [26].

While the concept of usability also involves the need for specialized skills or professional operation [39], the contactless nature of FMCW radars allows for remote monitoring, greatly reducing the interaction required from the user. This characteristic enhances acceptance and satisfaction by minimizing operational complexity and encouraging autonomous use.

5.3.2. Privacy

Privacy is another dimension of trust, because to achieve a truly trustworthy system, where the subject’s needs are considered as the center of the design of the technology, the minimization and anonymization of collected data are necessary. As outlined by [39], the concept of privacy is related to an individual’s right to be free from unwarranted intrusions into their personal life and affairs. Implementing telemonitoring systems for elderly individuals inevitably involves entering their private sphere. As expressed also by a recent review of AAL technologies [5], no device can be judged as completely private. However, in assessing the existing technological landscape, millimeter-wave sensors emerge as particularly advantageous, offering robust monitoring functions while imposing minimal disruption and limiting the extent of personal data collection.

One of the distinguishing features of mmWave radar technology in AAL lies in its inherent capability to support effective monitoring while limiting the collection of strictly unnecessary personal data. These systems operate by detecting physical parameters such as distance, velocity, and azimuth angle, without generating visual, audio, or biometric representations that could directly identify the individual. This focus on purely kinematic parameters enables high levels of privacy preservation, in accordance with the data minimization principles defined by the GDPR. The absence of sensitive content in the acquisition process reduces the risk of unauthorized disclosure and makes this technology particularly suitable for residential environments, where the perception of intrusiveness represents a critical barrier to the adoption of AAL solutions.

The degree of privacy protection provided by a given system largely depends on the quantity and sensitivity of the personal information it processes. In accordance with the European General Data Protection Regulation, personal data encompasses any information that can directly or indirectly identify an individual [40]. This includes a broad array of data types, which are considered “sensitive” when they reveal aspects such as sexual orientation, health records, political opinions, religious beliefs, ethnic origin, biometric features, or trade union affiliations [41].

From a technological standpoint, ensuring privacy involves designing systems capable of capturing and storing data while preserving the anonymity of the subject. A privacy-conscious approach aims at preventing the identification of individuals from collected data. In this context, mmWave radar systems have shown promise in providing effective surveillance and activity tracking with a reduced risk of intrusion on personal privacy, primarily because they can operate without necessitating extensive personal data disclosure. Such solutions facilitate a level of non-invasiveness that aligns with the long-term goal of safeguarding individual privacy while maintaining reliable monitoring.

Compared to other sensing modalities such as RGB/RGB-D cameras and audio-based systems [42,43], radar sensors are generally perceived as less intrusive in terms of privacy. This is mainly due to the fact that they capture physical parameters like distance, velocity, and angle, rather than detailed biometric or identifiable personal information. As a result, radar sensors can only detect aspects related to the subject’s movement, which are represented in the generated maps as simple spots or blobs, such as in Figure 6 and Figure 7, where a walking person appears merely as a point within the spatial map. Keeping the acquisitions anonymous by only detecting position and movement speed ensures respect for privacy, thus increasing the sense of safety in using technology that is not perceived as intrusive, generating trust in a device that can help them feel more at ease in their autonomy and independence.

However, recent research in mmWave sensing has explored radar micro-Doppler signatures for biometric and identification purposes. These micro-Doppler features, such as gait patterns or chest oscillations caused by the heartbeat, can be harnessed to authenticate and recognize individuals. Studies have demonstrated that the way a person walks or the pulsations of their chest, as captured through micro-Doppler data, are unique enough to serve as biometric identifiers, raising potential privacy concerns [10].

Further, radar micro-Doppler signatures have been explored for classifying personal attributes such as biological sex [44], based on the distinct movement characteristics observed within elderly cohorts. Age, too, can potentially be inferred from heartbeat micro-Doppler signals, although current research has primarily focused on pediatric populations under thirteen years old, where broad age classifications are feasible [45]. This progress highlights the capacity of radar systems not only for safety and surveillance but also for extracting sensitive personal features, thus emphasizing the importance of addressing privacy implications associated with these advanced capabilities.

In contrast, radar measurements seem to be unaffected by the skin color of the subject [29]. As noted in a recent review of the biomedical engineering applications of radar systems [46], there is a limited body of literature exploring the ethical implications of radar usage, especially regarding issues of privacy. As a result, addressing the ethical and privacy concerns associated with radar technology remains an important open challenge for future research.

5.3.3. Accuracy

As defined by ISO/IEC 5723:2022, accuracy refers to the degree of proximity between observed, computed, or estimated values and the actual or accepted true values [39]. It is one of the essential characteristics outlined by ISO for describing the reliability of a system. Accuracy serves as a crucial link between value and performance in radar systems, as it can be quantified through engineering-based metrics. Accuracy is a measure of performance, and it is fundamental in the definition of KPI; therefore, it has all the qualities previously mentioned, because it is quantifiable, verifiable, and continuously updated. Nevertheless, given that an accurate system is also a more trustworthy system, it is possible to reach a greater level of trustworthiness through short-term objectives of accuracy. This means that even though accuracy could change based on the application and the sensor, the final metric should reach a level that is satisfactory enough to achieve trust among the stakeholders.

The definition of accuracy may vary depending on the task for which it is employed; it is a widely used metric in machine learning, particularly for classification and regression tasks. For classification problems, accuracy is typically expressed as the ratio of correct predictions to the total number of predictions made. For instance, in the field of activity classification, fall detection, or walking pattern classification, it is possible to observe the definition of a number N of classes to discriminate; depending on the prediction of a developed model, a classification matrix can be obtained as in Figure 10.

Given a set of true positive (TP), true negative (TN), false positive (FP), and false negative (FN), it is possible to define the following metrics of performance:

$$\mathrm{Accuracy}={\displaystyle \frac{\mathrm{TP}+\mathrm{TN}}{\mathrm{TP}+\mathrm{TN}+\mathrm{FP}+\mathrm{FN}}}$$

(1)

Other performance metrics in the classification include precision, recall, and F1-score, as detailed in [38]:

$$\begin{array}{cc}\hfill \mathrm{Precision}& ={\displaystyle \frac{\mathrm{TP}}{\mathrm{TP}+\mathrm{FP}}}\hfill \end{array}$$

(2)

$$\begin{array}{cc}\hfill \mathrm{Recall}& ={\displaystyle \frac{\mathrm{TP}}{\mathrm{TP}+\mathrm{FN}}}\hfill \end{array}$$

(3)

$$\begin{array}{cc}\hfill \mathrm{F}1\mathrm{score}& =2·{\displaystyle \frac{\mathrm{Precision}\times \mathrm{Recall}}{\mathrm{Precision}+\mathrm{Recall}}}\hfill \end{array}$$

(4)

A recent review of common performance evaluation metrics used in machine learning applications for radar-based physiological signal regression illustrates the diversity of methods employed in the scientific literature to assess the efficacy of the radar system for regression tasks [10]. These metrics include, among others, the root mean square error, mean absolute error, correlation coefficient, and cosine similarity, each providing different insights into the accuracy and reliability of the regression outputs. When dealing with the regression of a particular physiological event, such as R-peak detections, it is common to choose a tolerance around the event, as shown in Figure 11.

The tolerance is needed to discriminate between false positives, false negatives, and true predictions. If the prediction is outside the defined tolerance mask, the prediction is false, the prediction is true if inside the mask, and there is a false negative prediction if no prediction can be bounded to the studied event. False positive and negative predictions and true predictions can be used to build the previous accuracy measure. Then, for the true predictions, it is possible to define time distance metrics as root mean square error and mean absolute error.

5.4. Human-Centric Design: Fairness

Elderly individuals remain vulnerable to discrimination and abuse, with recent data indicating that approximately 16% of those over 60 experienced some form of mistreatment within community settings in 2023, often perpetrated by caregivers or trusted persons [47]. Addressing this pressing issue, the World Health Organization (WHO) advocates for comprehensive strategies focused on prevention, intervention, and mitigation of the adverse effects associated with elder abuse. Technological support systems, such as the radar technology under discussion, are considered promising tools in this effort, as they have the potential to alleviate the caregiving burden and enhance protective measures.

Discrimination also presents a critical challenge within the field of information engineering, especially concerning the fairness of artificial intelligence (AI) algorithms. Recent research emphasizes that AI solutions can be vulnerable to bias, which may lead to unfair treatment of certain individuals or groups [48]. Discrimination, broadly defined as the “unfair treatment of an individual” [48], can arise when algorithms exhibit bias toward specific demographics. For example, Vilesov et al. [29] demonstrated that combining data from RGB sensors with a 77 GHz automotive radar improves both system performance and fairness in datasets including individuals with diverse skin tones. Their study underscores the advantage of radar sensors, which are less prone to skin-tone bias.

Furthermore, the concept of group fairness involves ensuring that the outcomes generated by algorithms are equitably distributed across different demographic groups within the dataset [48]. It is recognized that RGB sensors may be affected by variations in skin tone during the extraction of plethysmographic signals. However, integrating the analysis of color intensity variations with radar Doppler measurements has led to improvements in both accuracy and fairness metrics, thereby demonstrating a pathway toward more equitable and unbiased technological solutions.

5.5. Innovation for Sustainability: Economic and Environmental Aspects

The principle of innovation for sustainability is related to the technological advances in the field of AAL. The dimension of sustainability can be tackled from economic and environmental perspectives, which are briefly discussed in the next paragraphs.

The value of economic sustainability can be dissected into two critical components from both individual and societal viewpoints. The initial component pertains to the financial resources accessible to senior citizens, while the latter relates to the healthcare and social service expenses that governmental bodies must manage. Implementing a radar system for residential monitoring proves advantageous for both components: it caters to individuals who prefer to remain in their homes by offering technology that provides necessary monitoring without requiring a caregiver, which often incurs substantial costs. Furthermore, if patients can maintain autonomy, it presents an economic benefit and alleviates the demand for public healthcare facilities. By facilitating early detection of health issues and continuous surveillance, FMCW radars can diminish the frequency of hospitalizations and emergency room visits. Enabling individuals to reside in their homes for extended periods can lessen the necessity for long-term care facilities and the accompanying costs.

In relation to environmental sustainability, FMCW radars are engineered to be energy-efficient, thereby minimizing their ecological impact. Through advanced design and manufacturing techniques, these radars optimize material usage, thus reducing waste and enhancing sustainability.

6. Discussion and Open Challenges

Maintaining a stable quality of life is crucial, impacting not just the individuals themselves but also their interactions with others. Integrating radar technology as an additional home device supports personal independence and autonomy within a familiar and comfortable environment. This prevents the disruption of being displaced and maintains continuity in their usual settings. As a result, individuals can continue playing an active role in society, thereby improving their social participation

This study assessed how technology can generate significant value for societal benefit. Adopting a “value-oriented” analysis, we identified health and well-being, assistance and care, trustworthiness, fairness, sustainability, and social participation as the key values that millimeter-wave technology can foster. The potential applications of radar within the AAL sector, as discussed in this paper, are designed to enhance the autonomy of elderly individuals, helping them maintain independence and self-sufficiency, even in challenging situations.

For each identified value, we established corresponding key performances of radar systems, characterized by specific application domains and technical attributes where the technology demonstrates particular strength. The correlation between these key performances and their associated values is summarized in Figure 12.

In the domain of assistance and care, applications such as indoor human localization [26] and activity recognition through micro-Doppler characterization [25] are central. Similarly, the use of micro-Doppler analysis focused on the torso region for physiological signal sensing [38] relates directly to the value of health and well-being, enabling continuous monitoring of an individual’s health status.

Trustworthiness remains an area with limited research, primarily due to the recent development of relevant standards [39]. Ethical considerations, particularly regarding privacy, pose ongoing challenges, as scientific literature has shown limited engagement on this topic [46].

Compared to other sensing modalities such as video and audio, radar systems offer a significant advantage by providing better privacy preservation with minimal processing, thanks to their intrinsic environmental perception based on the Doppler effect.

Although some studies have attempted to extract personal data, such as sex [44] and age [45], or identify individuals within small groups [38] through micro-Doppler analysis, the ability to derive detailed personal information from raw Doppler data remains limited and is undoubtedly inferior to the richness of data obtainable from raw video and audio recordings. Additionally, research indicates that remote photoplethysmography extraction via video analysis, used to determine physiological indices, tends to be more susceptible to biases related to skin tone compared to radar-based methods [29]. Despite the increasing technical maturity of mmWave radar in controlled environments, real-world deployments in AAL contexts remain limited. Most studies rely on small-scale datasets or laboratory settings, which may not reflect the complexity and variability of actual residential care scenarios. As a result, the scalability, robustness, and user acceptance of radar-based systems in diverse real-life conditions are still largely untested. In parallel, ethical considerations surrounding radar use in AAL are often addressed in general terms, with limited attention to the specific nature of radiofrequency sensing. The unique ability of radar to operate non-invasively and anonymously is frequently cited as a privacy advantage, as shown in the coloumn ethical issues in the Table A2 in Appendix A, yet recent developments, such as biometric identification via micro-Doppler signatures, raise new ethical concerns. Unlike video or audio sensors, radar occupies an intermediate zone, where its perceived non-intrusiveness may lead to under regulation or ethical oversight. There is thus a pressing need for ethical frameworks tailored specifically to radar-based monitoring systems, especially in vulnerable populations.

One of the next steps for future research will be to define mathematical or algorithmic models that formalize the connection between signal-level radar outputs and high-level (KVIs). This will enable a more structured and quantifiable integration of ethical and technical considerations in the design process.

Another possible next step and one of the main future challenges could be to define and formalize quantitative metrics for KVIs, based on shared rules and mathematical models, so they can be directly considered during the design phase. This will be essential to promote more value-aware, measurable, and responsible technological development.

7. Conclusions

A multidisciplinary and value-centric approach, integrating technical innovation with ethical sensitivity and user perception, is necessary to develop adequate remote sensing technologies. In this study, we identified the key values that FMCW radar technology can promote within the context of AAL, including health and well-being, assistance and care, trustworthiness, fairness, sustainability, and social participation. A distinctive and innovative aspect of this work lies in adopting a value-driven design approach, where these values are not treated as secondary or resulting outcomes, but rather as primary design objectives. This approach enables a more responsible and user-centric technological development process by explicitly linking technical capabilities and application domains to the promotion of specific values. For each identified value, we analyzed the key performance of radar technology in terms of relevant technical attributes and applicability to use cases, highlighting its potential to support individual autonomy while preserving privacy and ensuring integration within the home environment. Only by treating values as core design drivers, rather than by-products, we can develop truly effective, acceptable, and sustainable solutions for long-term deployment. Future research should focus on three key directions: longitudinal in-home deployments of mmWave radar systems to assess long-term usability, impact on autonomy, and social engagement in real settings; the development of quantifiable models linking technical metrics to KVIs, enabling value-aware system optimization; and participatory validation studies involving caregivers and elderly users to ensure ethical acceptability and trust in practical applications.