Spatiotemporal Cleaning of PIR Sensor Data for Elderly Movement Monitoring
Abstract
:1. Introduction
2. Fundamental Concepts and Proposed Methodology
2.1. Definitions and Preliminaries
- PIR Sensor Network (): A graph where each vertex represents a PIR sensor positioned in the indoor environment and each edge represents a feasible direct line of movement between sensors that an individual may take.
- Detection Log (): A sequence of tuples where represents the sensor activated and is the timestamp when sensor was triggered.
- Activation Path of PIR sensors: A sequence of sensors such that, for each consecutive pair, there exists an edge, representing a possible movement path of a user through the PIR sensors to .
- Simulated User Trajectory Data from :
- ○
- Sensor Positions: Let and represent the spatial positions of the two PIR sensors, which are physically fixed within the indoor space.
- ○
- Activation Timing: Let and be the respective timestamps when sensors and are activated.
- Definition of User Presence Points: The user’s presence point at any given time between two sensor activations can be defined based on a linear interpolation between the positions of the two sensors, using the sampling rate as follows:
- ○
- Hop Count Function: The hop count between two sensors and , denoted , calculates the minimum number of edges in the shortest path between them in the graph .
- ○
- Distance Function: The physical distance between two sensors and , denoted , is determined by their spatial positions and the setup during sensor installation.
- ○
- Speed Calculation: The walking speed between two detections and , is calculated as
2.2. False Positives and False Negatives
- False Positive (FP): A detection by the PIR sensor that does not correspond to actual human movement.
- ○
- Definition: For a detection event , it is considered a false positive if the sensor is activated but no movement is detected by any adjacent sensors within a specific time threshold .
- ○
- Condition: is activated and , if , then is not activated.
- False Negative (): A missed detection where the PIR sensor fails to detect actual movement.
- ○
- Definition: For a sequence of detections and , a false negative occurs if the subject moves between and but no intermediate detections are recorded by sensors located along the path between these two sensors.
- ○
- Condition: is activated, is activated and such that is not activated within []. Here, represents the set of sensors located along the direct movement path between and . A false negative is detected when at least one intermediate sensor is expected to detect the movement but fails to activate during the subject’s movement between and .
2.3. Subject Movement Modes: ‘Moving’ and ‘Staying’
- : The subject is detected transitioning between sensors without significant pauses.
- ○
- Condition: , where is the minimum threshold for movement speed, ensuring the subject is actively transitioning between sensors. This mode is triggered when the subject’s movement speed exceeds , indicating continuous movement.
- : The subject remains near a sensor for an extended period, indicating a pause or rest.
- ○
- Condition: where is a time threshold indicating how long the subject remains near the sensor and the speed between the two activations is close to zero. This mode is used when the subject lingers or rests near a sensor for a significant period.
2.4. Integration of Non-Deterministic Tracking and Late-Binding Adjustment for Path Monitoring
2.5. Path Tracking Using Non-Deterministic Tracking (NDT)
- (1)
- Initialization: Begin with the initial detection log ().
- (2)
- Path Exploration: For each initial detection , evaluate all subsequent detections where .
- (3)
- Feasibility Check:
- ○
- Calculate the hop count Hop and .
- ○
- Retain the path if it satisfies the following conditions:
- ■
- Hop count.
- ■
- Speed threshold
(e.g., ).
- (4)
- False Positive Check: If the conditions are not met and no adjacent sensors detect movement, mark the detection as a false positive.
- (5)
- Mode Identification: Classify movement as ‘Moving‘ if the speed exceeds a threshold, or ‘Staying‘ if the time between activations is long and the speed is near zero.
- (6)
- Branching: Generate multiple branches of the detection log, each representing a different possible correction, with the movement mode labeled for each path.
2.6. Path Tracking Using Late-Binding Adjustment (LBA)
- (1)
- Initial Log Processing: Begin by creating a copy of the detection log , where each detection is assigned an initial state.
- (2)
- Detection Iteration: For each detection event , evaluate all subsequent detection events within the log.
- (3)
- Contradiction Identification:
- ○
- Calculate the speed () and the hop count (=Hop) between each detection pair .
- ○
- If the calculated speed exceeds realistic thresholds or if the hop count is abnormally high, mark the initial detection as potentially erroneous.
- ○
- Additionally, check for false negatives by identifying missing intermediate detections expected between and .
- (4)
- State Adjustment:
- ○
- Adjust the log based on the last confirmed correct detection, modifying states and timestamps for detections from to .
- ○
- Label ‘Staying‘ if the subject stayed near a sensor for an extended period without moving.
- ○
- Update false positives if no corresponding movement is confirmed by adjacent sensors.
Algorithm 1: The data cleaning algorithm that integrates NDT and LBA |
Input: DetectionLog DL, SensorGraph G(V, E), TimeThreshold , HopThreshold Hop_Max Input: Speed Threshold Speed_Max Output: RefinedPath RP 1 Initialize HypothesesList 2 For each detection event in DL 3 a. Create initial path ] 4 b. Add P to H 5 For each hypothesis P in H: 6 a. While there are subsequent detections : 7 i. If AND : 8 Append to P 9 ii. Else If no valid adjacent sensor : 10 Mark as False Positive (FP) 11 Break 12 Apply Late Binding Adjustment (LBA) to refine hypotheses in H: 13 a. For each hypothesis P in H: 14 i. Backtrack through P to correct inconsistencies 15 ii. If a detection is invalid: 16 Remove as FP or interpolate as FN 17 iii. Optimize path based on updated data 18 Consolidate valid hypotheses to form RefinedPath RP 19 Return RP |
3. Background and Experimental System: Hardware, Software, and Field Configuration
3.1. Detection Range and Error Reduction in Passive Infrared (PIR) Sensors
3.1.1. Limitation of the PIR Sensor’s Detection Range
3.1.2. Construction of PIR Sensors Module
3.2. Modeling and Evaluation of PIR Sensor-Based Walking Speed and Movement Tracking in Elderly Environments
3.2.1. Modeling of Sensor States and Transfers Between Sensors
3.2.2. Sensor Time Series Data
- {‘DateTime’: ‘2024/05/01 16:37:18.900000’, ‘EpochTime’: ‘1714549038.900’, ‘ID’:’ PIRS01’, ‘Status’: ‘ON’}
3.2.3. Elderly Housing Model and Sensor Placement
4. Implementation of Spatiotemporal Data Cleaning Method for PIR Sensor Data
4.1. Data Cleaning Method Using Network Topology and Subjects’ Walking Speed
4.2. Implementing Late Binding to Refine Trajectory Data Corrections
5. Evaluation of Spatiotemporal Cleaning of PIR Sensor Data by Simulation
- (1)
- Integrated Simulation Data Generator (ISDG), which creates PIR sensor data reflecting the subject’s movements within rooms. To evaluate the proposed method, it is possible to introduce various error patterns, such as false positives and missed detections, which are difficult to replicate in real-world experiments.
- (2)
- LBIS data cleaning filter, which remove false positives and other irrelevant information from PIRS time series data produced by the ISDG. Implements LBIS in software to perform data cleaning on sensor data containing errors. This process is also used in real-world experiments.
- (3)
- SensorPathFinder, which visualizes the data cleaned by the LBIS. To evaluate the data before and after the data cleaning process, the tracking results are visualized.
5.1. Integrated Simulation Data Generator (ISDG)
5.2. Spatiotempral Data Cleaning Filter
- Zero IDs: No change from the previous state; the state at time is copied to the extended PIRS state change table.
- One ID: Indicates potential subject detection, initiating movement determination. If no provisional subject exists, it is evaluated whether movement from the last detected location to is feasible. If a provisional subject is present and can move to , other provisional subjects are removed, and updates are made retrospectively to the state change table. If movement is not feasible, is flagged as a false positive.
- Two or more IDs: Uncertain movement decision: ‘(, ‘A?’)’ is added as a provisional subject to the extended PIRS state change table.
5.3. Software Development for Converting PIR Sensor Log Data into Trajectory Data
5.4. Simulation Scenarios of Error Data Generation from PIRS Units and Their Corrections
- Case 1: False positive detection by a PIRS unit far from the subject.
- Case 2: Subject reversal during PIRS blocking time.
- Case 3: Simultaneous detection by two PIRS units.
5.5. Evaluation of the Conversion of PIR Sensor Log Data into Trajectory Data
5.5.1. Case 1: False Positive Detection by a PIRS Unit at a Distance from the Subject
5.5.2. Case 2: Subject Reversal During PIRS Blocking Time
5.5.3. Case 3: Simultaneous Detection by Two PIRS Units
- Naive method (Figure 23a): The trajectory starts at , which has the smallest ID, and then moves to . No path exists between and due to simultaneous detections, suggesting the subject moves to immediately upon arriving at .
- Network topology and walking speed (Figure 23b): This analysis chooses over , both detected simultaneously, due to its smaller ID. is deemed a false positive because there is no time lapse between the subject moving from . The movement from to then follows a two-hop path via , reflecting the absence of a direct link in the network topology.
- Late-binding approach (Figure 23c): This method refines the tracking by correcting discrepancies observed in simpler methods, aligning the trajectory closely with the actual movement path.
6. Evaluation Using Real-World Experiments
6.1. Real-World Experiment Setup and Participants for Spatiotemporal Cleaning of PIR Sensor Data
6.2. Results from Real-World Testing
6.3. Comparative Analysis: Simulation vs. Real-World Results
7. Discussion
8. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- World Health Organization. Aging and health. Available online: https://www.who.int/news-room/fact-sheets/detail/ageing-and-health (accessed on 29 February 2024).
- Enssle, F.; Kabisch, N. Urban green spaces for the social interaction, health and well-being of older people—An integrated view of urban ecosystem services and socio-environmental justice. Environ. Sci. Policy 2020, 109, 36–44. [Google Scholar] [CrossRef]
- Uddin, M.Z.; Khaksar, W.; Torresen, J. Ambient Sensors for Elderly Care and Independent Living: A Survey. Sensors 2018, 18, 2027. [Google Scholar] [CrossRef] [PubMed]
- Tomii, S.; Ohtsuki, T. Falling detection using multiple doppler sensors. In Proceedings of the 2012 IEEE 14th International Conference on e-Health Networking, Applications and Services (Healthcom), Beijing, China, 10–13 October 2012; pp. 196–201. [Google Scholar] [CrossRef]
- Rajavel, R.; Ravichandran, S.K.; Harimoorthy, K.; Nagappan, P.; Gobichettipalayam, K.R. IoT-based smart healthcare video surveillance system using edge computing. J. Ambient Intell. Humaniz. Comput. 2022, 13, 3195–3207. [Google Scholar] [CrossRef]
- Jo, T.H.; Ma, J.H.; Cha, S.H. Elderly Perception on the Internet of Things-Based Integrated Smart-Home System. Sensors 2021, 21, 1284. [Google Scholar] [CrossRef]
- Shah, S.A.; Ahmad, J.; Masood, F.; Shah, S.Y.; Pervaiz, H.; Taylor, W.; Imran, M.A.; Abbasi, Q.H. Privacy-Preserving Wandering Behavior Sensing in Dementia Patients Using Modified Logistic and Dynamic Newton Leipnik Maps. IEEE Sens. J. 2021, 21, 3669–3679. [Google Scholar] [CrossRef]
- Sugianto, N.; Tjondronegoro, D.; Stockdale, R.; Yuwono, E.I. Privacy-preserving AI-enabled video surveillance for social distancing: Responsible design and deployment for pbulci spaces. Inf. Technol. People 2024, 37, 998–1022. [Google Scholar] [CrossRef]
- Wu, C.M.; Chen, X.Y.; Wen, C.Y.; Sethares, W.A. Cooperative Network PIR Detection System for Indoor Human Localization. Sensors 2021, 21, 6180. [Google Scholar] [CrossRef] [PubMed]
- Tabbakha, N.E.; Ooi, C.-P.; Tan, W.-H.; Tan, Y.F. A wearable device for machine learning based elderly’s activity tracking and indoor location system. Bull. Electr. Eng. Inform. 2021, 10, 927–939. [Google Scholar] [CrossRef]
- Ngamakeur, K.; Yongchareon, S.; Yu, J.; Islam, S. Passive infrared sensor dataset and deep learning models for device-free indoor localization and tracking. Pervasive Mob. Comput. 2023, 88, 101721. [Google Scholar] [CrossRef]
- Tao, S.; Kudo, M.; Pei, B.N.; Nonaka, H.; Toyama, J. Multiperson Location and Their Soft Tracking in a Binary Infrared Sensor Network. IEEE Trans. Hum. Mach. Syst. 2015, 45, 550–560. [Google Scholar] [CrossRef]
- Veronese, F.; Comai, S.; Mangano, S.; Matteucci, M.; Salice, F. PIR Probability Model for a Cost/Reliability Tradeoff Unobtrusive Indoor Monitoring System, Lecture Notes of the Institute for Computer Sciences. In Proceedings of the International Conference on Smart Objects and Technologies for Social Good, Pisa, Italy, 29–30 November 2017; p. 61. [Google Scholar] [CrossRef]
- Shokrollahi, A.; Persson, J.A.; Malekian, R.; Sarkheyli-Hägele, A.; Karlsson, F. Passive Infrared Sensor-Based Occupancy Monitoring in Smart Buildings: A Review of Methodologies and Machine Learning Approaches. Sensors 2024, 24, 1533. [Google Scholar] [CrossRef] [PubMed]
- Khan, D.S.; Kolarik, J.; Hviid, C.A.; Weitzmann, P. Method for long-term mapping of occupancy patterns in open-plan and single office spaces by using passive-infrared (PIR) sensors mounted below desks. Energy Build. 2021, 230, 110534. [Google Scholar] [CrossRef]
- Sultanabanu, K.; Liyakat, S.; Kazi, K.; Kosgiker, G.; Gund, V. PIR Sensor-Based Arduino Home Security System. Int. J. Eng. Res. Technol. 2019, 8. [Google Scholar]
- Masciadri, A.; Lin, C.; Comai, S.; Salice, F. A Multi-Resident Number Estimation Method for Smart Homes. Sensors 2022, 22, 4823. [Google Scholar] [CrossRef] [PubMed]
- Narayana, S.; Prasad, R.V.; Rao, V.S.; Prabhakar, T.V.; Kowshik, S.S.; Iyer, M.S. PIR Sensors: Characterization and Novel Localization Technique. In Proceedings of the 14th International Conference on Information Processing in Sensor Networks, Seattle, WA, USA, 13–16 April 2015; pp. 142–153. [Google Scholar] [CrossRef]
- Sakai, H.; Okuma, H.; Wu, M.; Nakata, M. Rough Non-Deterministic Information. In The Handbook on Reasoning-Based Intelligent Systems; World Scientific Publishing Co Pte Ltd.: Singapore, Singapore, 2013; pp. 81–118. [Google Scholar] [CrossRef]
- Sakai, H.; Nakata, M. On Possible Rules and Apriori Algorithm in Non-deterministic Information Systems. In Rough Sets and Current Trends in Computing; RSCTC 2006—Lecture Notes in Computer Science; Greco, S., Ed.; Springer: Berlin/Heidelberg, Germany, 2006; Volume 4259. [Google Scholar] [CrossRef]
- Taylor, J.T.; Taylor, W.T. More About Late Binding. In Patterns in the Machine; Apress: Berkeley, CA, USA, 2021. [Google Scholar] [CrossRef]
- Pontelli, E.; Ranjan, D.; Gupta, G. The Complexity of Late-Binding in Dynamic Object-Oriented Languages. In Principles of Declarative Programming, ALP PLILP 1998—Lecture Notes in Computer Science; Palamidessi, C., Glaser, H., Meinke, K., Eds.; Springer: Berlin/Heidelberg, Germany, 1998; Volume 1490. [Google Scholar] [CrossRef]
- Utsumi, T.; Arikawa, M.; Hashimoto, M. MUS3E: A Mobility Ubiquitous Sensor Edge Environment for the Elderly. Electronics 2023, 12, 3003. [Google Scholar] [CrossRef]
- Akizuki Denshi Tsusho. SB612B Datasheet. Available online: https://akizukidenshi.com/goodsaffix/SB612B_20221128.pdf (accessed on 5 March 2024).
- ROBU.IN. Pyroelectric Infrared Radial Sensor. Available online: https://robu.in/wp-content/uploads/2020/09/Pir-BM612.pdf (accessed on 5 March 2024).
- Yoshimura, N.; Muraki, S.; Oka, H.; Tanaka, S.; Ogata, T.; Kawaguchi, H.; Akune, T.; Nakamura, K. Association between new indices in the locomotive syndrome risk test and decline in mobility: Third survey of the ROAD study. J. Orthop. Sci. 2015, 20, 896–905. [Google Scholar] [CrossRef]
- Hirai, T.; Chidori, K.; Watanabe, N.; Hoshino, M.; Kawai, H.; Ueno, N.; Inoue, D.; Harada, T.; Maki, K.; Shimoto, T. Evaluation of the Grading Ability of Walking Speed by the Elderly. J. Phys. Ther. Sci. 2008, 23, 711–715. [Google Scholar] [CrossRef]
Items | Values |
---|---|
Supply voltage | DC 3.3–12 V |
Rated detection distance | 8 m (max) |
Horizontal detection range | 115 degrees |
Detecting zone | 64 zones |
Blocking time | 2 s |
Price | 4 USD |
Subject | Methods | Average Error (Meters) | Maximum Error (Meters) | Minimum Error (Meters) | Standard Deviation |
---|---|---|---|---|---|
A | Naive without network topology | 2.0 | 9.9 | 0.1 | 1.82 |
Naive with network topology | 1.7 | 9.5 | 0.1 | 1.83 | |
Constrained to a maximum hop count of two or less and a walking speed (2.0 m/s) | 1.8 | 6.6 | 0.1 | 1.77 | |
Late binding | 1.0 | 3.1 | 0.1 | 0.79 | |
B | Naive without network topology | 1.1 | 9.0 | 0.1 | 1.4 |
Naive with network topology | 1.1 | 8.0 | 0.1 | 1.3 | |
Constrained to a maximum hop count of two or less and a walking speed (2.0 m/s) | 1.2 | 4.2 | 0.1 | 0.9 | |
Late binding | 1.0 | 2.6 | 0.1 | 0.9 | |
C | Naive without network topology | 3.0 | 9.9 | 0.0 | 2.27 |
Naive with network topology | 2.9 | 10.1 | 0.0 | 2.22 | |
Constrained to a maximum hop count of two or less and a walking speed (2.0 m/s) | 1.9 | 6.6 | 0.0 | 1.8 | |
Late binding | 1.3 | 6.6 | 0.1 | 1.35 |
Subject | False Positive (FP) | False Negative (FN) | The Number of False Positive Corrections | The Number of False Negative Corrections |
---|---|---|---|---|
A | 6 | 3 | 5 | 3 |
B | 8 | 3 | 8 | 2 |
C | 16 | 1 | 16 | 1 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Utsumi, T.; Arikawa, M. Spatiotemporal Cleaning of PIR Sensor Data for Elderly Movement Monitoring. Electronics 2024, 13, 4707. https://doi.org/10.3390/electronics13234707
Utsumi T, Arikawa M. Spatiotemporal Cleaning of PIR Sensor Data for Elderly Movement Monitoring. Electronics. 2024; 13(23):4707. https://doi.org/10.3390/electronics13234707
Chicago/Turabian StyleUtsumi, Tomihiro, and Masatoshi Arikawa. 2024. "Spatiotemporal Cleaning of PIR Sensor Data for Elderly Movement Monitoring" Electronics 13, no. 23: 4707. https://doi.org/10.3390/electronics13234707
APA StyleUtsumi, T., & Arikawa, M. (2024). Spatiotemporal Cleaning of PIR Sensor Data for Elderly Movement Monitoring. Electronics, 13(23), 4707. https://doi.org/10.3390/electronics13234707