Special Issue "Smart Sensing for Pervasive Health"
A special issue of Sensors (ISSN 1424-8220). This special issue belongs to the section "Wearables".
Deadline for manuscript submissions: 30 June 2023 | Viewed by 2355
Special Issue Editors

Interests: artificial intelligence; machine learning; wearable computing; intelligent systems; activity recognition; time series analysis
Special Issues, Collections and Topics in MDPI journals

Interests: pervasive health; interactive systems; human–computer interaction; virtual coaching; privacy-by-design

Interests: ambient intelligence; interpretation of sensor data; application of AI in healthcare; machine learning; decision support
Special Issues, Collections and Topics in MDPI journals

Interests: human–computer interaction; pervasive health; mobile and wearable computing; accessibility

Interests: pervasive health; Internet-of-Health-Things; wearable computing
Special Issues, Collections and Topics in MDPI journals
Special Issue Information
Dear Colleagues,
Recent advances on sensor technologies and their pervasiveness allow the development of numerous applications and services to improve people’s health. This Special Issue invites papers presenting novel research on utilizing sensors placed on users’ bodies and in their environments to understand and improve their health and wellbeing. We are interested in sensing devices, methods, and applications for obtaining and analyzing different categories of data (physiological, behavioral, emotional, environmental, etc.), and utilizing different technologies such as signal processing, machine learning, etc. for empowering citizens and health professionals to take action based on the analyzed sensor data.
This Special Issue is supported by the WideHealth Project, which is a European TWINNING project (No. 952279) that supports dissemination activities on the topics of eHealth and Pervasive health, and aims to enable a new generation of researchers on the aforementioned topics.
Dr. Hristijan Gjoreski
Dr. Oscar Mayora
Dr. Mitja Luštrek
Dr. Tiago Guerreiro
Prof. Dr. Bert Arnrich
Guest Editors
Manuscript Submission Information
Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.
Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Sensors is an international peer-reviewed open access semimonthly journal published by MDPI.
Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.
Keywords
- smart sensing
- pervasive health
- wearables
- machine learning
- deep learning
- sensor fusion
- human activities and behavior analytics
- physiological sensing
Planned Papers
The below list represents only planned manuscripts. Some of these manuscripts have not been received by the Editorial Office yet. Papers submitted to MDPI journals are subject to peer-review.
Planned Paper1#
Type of Paper: Research Article
Tentative Title: Walking identification in aged residential care: a computationally inexpensive and orientation independent algorithm with a single 3-axis accelerometer on the trunk
Authors: Mhairi K. MacLean1, Rana Zia Ur Rehman1, Ngaire Kerse2, Lynne Taylor2, Lynn Rochester1, Silvia Del Din1
Affiliations:1.Newcastle University, Translational and Clinical Research Institute; 2.University of Auckland, Faculty of Medical and Health Sciences
Abstract:
Background and Aims: Measurement of walking activity in long-term residential care environments is important for understanding mobility, assessing the effects of interventions, tracking disease progression, and identifying unmet mobility needs . Previous research has shown that gait classification algorithms can identify multiple gait characteristics like walking speed, gait asymmetry, and stride length with a single 3-axis inertial measurement unit (IMU) worn on the lower back. Gait classification in a long-term residential care environment is challenging due to slow walking speeds, halting gait patterns, and reliance on walking aids which alter gait characteristics. The aim of this study was to develop a computationally inexpensive algorithm to classify gait in a residential care environment. To improve robustness of the gait classifier, we aimed to make the algorithm independent of the IMU orientation, ensuring that improper placement of the IMU would not lead to significant data loss.
Methods: Data were collected from 27 ambulatory residents in long-term care. A clinician secured the IMU to the fifth lumbar vertebrae with a hydrogel adhesive. Participants were asked to walk around the care home, stand, and sit during a 5–15-minute period. The researcher recorded the order of activities and time taken for each activity, with sit-stand and stand-sit transfers counting as activities. We also recorded a video of the participants feet during the data collection. The IMU sampled data at 100 Hz.
Acceleration data were first adjusted by subtracting the low pass filtered data, we then transformed all 3 linear accelerations into a single magnitude signal. Thereafter, we classified each timepoint as walking or not-walking using thresholds, and removed small gaps in activity with a smoothing technique. We excluded walking bouts less than 2 seconds. We compared the algorithm classification to the clinician labelled data at each timepoint to find accuracy, sensitivity, precision, specificity, and F1 score. We averaged these scores across participants.
Results: Our algorithm performed well with an average accuracy of 79% ±19 standard deviation (s.d.). Sensitivity was also 79% (±27 s.d). Precision was 83% ±25, specificity was 82% ±11, and F1 score was 80% ±26 (mean±s.d).
Discussion and Conclusions.We validated a computationally inexpensive and orientation independent algorithm for classifying gait in a long-term care environment. Our results show the algorithm worked relatively well across participants. The importance of an easy to use, computationally inexpensive technique to classify gait was accentuated by the ability of our algorithm to identify bouts of walking that were not labelled as such during the data collection, although the video data exhibits walking.
Planned Paper2#
Title: "Recognising activities using ear-worn sensors and machine learning" Authors:Matias Laporte,Davide Casnici, Martin Gjoreski,Shkurta Gashi, Silvia Santini, Marc Langheinrich Affiliation: Università della Svizzera italiana
Abstract: Working toward context recognition for human memory augmentation systems, we developed Human Activity Recognition (HAR) machine learning pipeline using data from earable devices. In this paper, we analyze how the earables can be used to detect different types of verbal (e.g., speaking) and non-verbal (e.g., head shaking) interactions and other activities (e.g., standing still). We collected a dataset of ear-located IMU sensors from 30 participants and compared classical machine learning methods with state-of-the-art deep learning models to classify the raw data into a set of predefined activities. We explored how different parameters -- including sampling frequency, window size, and type of sensors -- influence the models’ performance. The best-performing model in the leave-one-subject-out evaluation, a spectro-temporal residual network (STResNet), achieved an F1-score of 0.69 in recognizing seven different activities: nodding, speaking, eating, staying, head shaking, walking, and walking while speaking.
Keywords: earable computing, human activity recognition, machine learning, deep learning, human memory augmentation