2024
Matey-Sanz, Miguel
Human Activity Recognition with Consumer Devices and Real-Life Perspectives PhD Thesis
2024.
Abstract | Links | BibTeX | Tags: activity recognition, machine learning, smartphone app, smartwatch
@phdthesis{Matey2024c,
title = {Human Activity Recognition with Consumer Devices and Real-Life Perspectives},
author = {Miguel Matey-Sanz},
doi = {http://dx.doi.org/10.6035/14101.2024.663821},
year = {2024},
date = {2024-10-30},
abstract = {During the last decade, research on human activity recognition has grown due to its applications in diverse fields such as video surveillance, exercise monitoring or health monitoring systems. In the latter case, researchers are putting their efforts into using human activity recognition in monitoring elderly people, for example, for fall prevention and detection applications. Existing research usually has drawbacks regarding their requirements regarding sensing devices (e.g., cost, quantity, location). Therefore, research needs to keep these drawbacks in mind to have a real impact on society. This thesis addresses the abovementioned issue by focusing on the feasibility of the use of consumer devices such as smartphones and smartwatches, and cheap devices like microcontrollers, for human activity recognition and its application in real-life problems.},
keywords = {activity recognition, machine learning, smartphone app, smartwatch},
pubstate = {published},
tppubtype = {phdthesis}
}
During the last decade, research on human activity recognition has grown due to its applications in diverse fields such as video surveillance, exercise monitoring or health monitoring systems. In the latter case, researchers are putting their efforts into using human activity recognition in monitoring elderly people, for example, for fall prevention and detection applications. Existing research usually has drawbacks regarding their requirements regarding sensing devices (e.g., cost, quantity, location). Therefore, research needs to keep these drawbacks in mind to have a real impact on society. This thesis addresses the abovementioned issue by focusing on the feasibility of the use of consumer devices such as smartphones and smartwatches, and cheap devices like microcontrollers, for human activity recognition and its application in real-life problems.
Matey-Sanz, Miguel; González-Pérez, Alberto; Casteleyn, Sven; Granell-Canut, Carlos
Implementing and Evaluating the Timed Up and Go Test Automation Using Smartphones and Smartwatches Journal Article
In: IEEE Journal of Biomedical and Health Informatics, vol. 28, iss. 11, pp. 6594 - 6605, 2024, ISSN: 2168-2208.
Abstract | Links | BibTeX | Tags: activity recognition, machine learning, Mobile apps, symptoms, wearables
@article{Matey2024b,
title = {Implementing and Evaluating the Timed Up and Go Test Automation Using Smartphones and Smartwatches},
author = {Miguel Matey-Sanz and Alberto González-Pérez and Sven Casteleyn and Carlos Granell-Canut},
doi = {https://doi.org/10.1109/JBHI.2024.3456169},
issn = {2168-2208},
year = {2024},
date = {2024-09-09},
urldate = {2024-09-09},
journal = {IEEE Journal of Biomedical and Health Informatics},
volume = {28},
issue = {11},
pages = {6594 - 6605},
abstract = {Physical performance tests aim to assess the physical abilities and mobility skills of individuals for various healthcare purposes. They are often driven by experts and usually performed at their practice, and therefore they are resource-intensive and time-demanding. For tests based on objective measurements (e.g., duration, repetitions), technology can be used to automate them, allowing the patients to perform the test themselves, more frequently and anywhere, while alleviating the expert from supervising the test. The well-known Timed Up and Go (TUG) test, typically used for mobility assessment, is an ideal candidate for automation, as inertial sensors (among others) can be deployed to detect the various movements constituting the test without expert supervision. To move from expert-led testing to self-administered testing, we present a mHealth system capable of automating the TUG test using a pocket-sized smartphone or a wrist smartwatch paired with a smartphone, where data from inertial sensors are used to detect the activities carried out by the patient while performing the test and compute their results in real time. All processing (i.e., data processing, machine learning-based activity inference, results calculation) takes place on the smartphone. The use of both devices to automate the TUG test was evaluated (w.r.t. accuracy, reliability and battery consumption) and mutually compared, and set off with a reference method, obtaining excellent Bland-Altman agreement results and Intraclass Correlation Coefficient reliability. Results also suggest that the smartwatch-based system performs better than the smartphone-based system.},
keywords = {activity recognition, machine learning, Mobile apps, symptoms, wearables},
pubstate = {published},
tppubtype = {article}
}
Physical performance tests aim to assess the physical abilities and mobility skills of individuals for various healthcare purposes. They are often driven by experts and usually performed at their practice, and therefore they are resource-intensive and time-demanding. For tests based on objective measurements (e.g., duration, repetitions), technology can be used to automate them, allowing the patients to perform the test themselves, more frequently and anywhere, while alleviating the expert from supervising the test. The well-known Timed Up and Go (TUG) test, typically used for mobility assessment, is an ideal candidate for automation, as inertial sensors (among others) can be deployed to detect the various movements constituting the test without expert supervision. To move from expert-led testing to self-administered testing, we present a mHealth system capable of automating the TUG test using a pocket-sized smartphone or a wrist smartwatch paired with a smartphone, where data from inertial sensors are used to detect the activities carried out by the patient while performing the test and compute their results in real time. All processing (i.e., data processing, machine learning-based activity inference, results calculation) takes place on the smartphone. The use of both devices to automate the TUG test was evaluated (w.r.t. accuracy, reliability and battery consumption) and mutually compared, and set off with a reference method, obtaining excellent Bland-Altman agreement results and Intraclass Correlation Coefficient reliability. Results also suggest that the smartwatch-based system performs better than the smartphone-based system.
2023
Matey-Sanz, Miguel; Casteleyn, Sven; Granell-Canut, Carlos
Dataset of inertial measurements of smartphones and smartwatches for human activity recognition Journal Article
In: Data in Brief, vol. 51, pp. 109809, 2023, ISSN: 2352-3409.
Abstract | Links | BibTeX | Tags: activity recognition, dataset, machine learning, smartphone app, smartwatch, symptoms
@article{Matey2023c,
title = {Dataset of inertial measurements of smartphones and smartwatches for human activity recognition},
author = {Miguel Matey-Sanz and Sven Casteleyn and Carlos Granell-Canut},
doi = {https://doi.org/10.1016/j.dib.2023.109809},
issn = {2352-3409},
year = {2023},
date = {2023-12-15},
journal = {Data in Brief},
volume = {51},
pages = {109809},
abstract = {This article describes a dataset for human activity recognition with inertial measurements, i.e., accelerometer and gyroscope, from a smartphone and a smartwatch placed in the left pocket and on the left wrist, respectively. Twenty-three heterogeneous subjects (μ = 44.3, σ = 14.3, 56% male) participated in the data collection, which consisted of performing five activities (seated, standing up, walking, turning, and sitting down) arranged in a specific sequence (corresponding with the TUG test). Subjects performed the sequence of activities multiple times while the devices collected inertial data at 100 Hz and were video-recorded by a researcher for data labelling purposes. The goal of this dataset is to provide smartphone- and smartwatch-based inertial data for human activity recognition collected from a heterogeneous (i.e., age-diverse, gender-balanced) set of subjects. Along with the dataset, the repository includes demographic information (age, gender), information about each sequence of activities (smartphone's orientation in the pocket, direction of turns), and a Python package with utility functions (data loading, visualization, etc). The dataset can be reused for different purposes in the field of human activity recognition, from cross-subject evaluation to comparison of recognition performance using data from smartphones and smartwatches.},
keywords = {activity recognition, dataset, machine learning, smartphone app, smartwatch, symptoms},
pubstate = {published},
tppubtype = {article}
}
This article describes a dataset for human activity recognition with inertial measurements, i.e., accelerometer and gyroscope, from a smartphone and a smartwatch placed in the left pocket and on the left wrist, respectively. Twenty-three heterogeneous subjects (μ = 44.3, σ = 14.3, 56% male) participated in the data collection, which consisted of performing five activities (seated, standing up, walking, turning, and sitting down) arranged in a specific sequence (corresponding with the TUG test). Subjects performed the sequence of activities multiple times while the devices collected inertial data at 100 Hz and were video-recorded by a researcher for data labelling purposes. The goal of this dataset is to provide smartphone- and smartwatch-based inertial data for human activity recognition collected from a heterogeneous (i.e., age-diverse, gender-balanced) set of subjects. Along with the dataset, the repository includes demographic information (age, gender), information about each sequence of activities (smartphone's orientation in the pocket, direction of turns), and a Python package with utility functions (data loading, visualization, etc). The dataset can be reused for different purposes in the field of human activity recognition, from cross-subject evaluation to comparison of recognition performance using data from smartphones and smartwatches.
Matey-Sanz, Miguel; Torres-Sospedra, Joaquín; González-Pérez, Alberto; Casteleyn, Sven; Granell-Canut, Carlos
Analysis and Impact of Training Set Size in Cross-Subject Human Activity Recognition Proceedings Article
In: Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, pp. 391–405, Springer, Cham, 2023, ISBN: 978-3-031-49018-7.
Abstract | Links | BibTeX | Tags: activity recognition, machine learning, smartphone app, smartwatch, symptoms
@inproceedings{Matey2023b,
title = {Analysis and Impact of Training Set Size in Cross-Subject Human Activity Recognition},
author = {Miguel Matey-Sanz and Joaquín Torres-Sospedra and Alberto González-Pérez and Sven Casteleyn and Carlos Granell-Canut},
doi = {https://doi.org/10.1007/978-3-031-49018-7_28},
isbn = {978-3-031-49018-7},
year = {2023},
date = {2023-12-01},
booktitle = {Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications},
volume = {14469},
pages = {391–405},
publisher = {Springer, Cham},
series = {Lecture Notes in Computer Science},
abstract = {The ubiquity of consumer devices with sensing and computational capabilities, such as smartphones and smartwatches, has increased interest in their use in human activity recognition for healthcare monitoring applications, among others. When developing such a system, researchers rely on input data to train recognition models. In the absence of openly available datasets that meet the model requirements, researchers face a hard and time-consuming process to decide which sensing device to use or how much data needs to be collected. In this paper, we explore the effect of the amount of training data on the performance (i.e., classification accuracy and activity-wise F1-scores) of a CNN model by performing an incremental cross-subject evaluation using data collected from a consumer smartphone and smartwatch. Systematically studying the incremental inclusion of subject data from a set of 22 training subjects, the results show that the model’s performance initially improves significantly with each addition, yet this improvement slows down the larger the number of included subjects. We compare the performance of models based on smartphone and smartwatch data. The latter option is significantly better with smaller sizes of training data, while the former outperforms with larger amounts of training data. In addition, gait-related activities show significantly better results with smartphone-collected data, while non-gait-related activities, such as standing up or sitting down, were better recognized with smartwatch-collected data.},
keywords = {activity recognition, machine learning, smartphone app, smartwatch, symptoms},
pubstate = {published},
tppubtype = {inproceedings}
}
The ubiquity of consumer devices with sensing and computational capabilities, such as smartphones and smartwatches, has increased interest in their use in human activity recognition for healthcare monitoring applications, among others. When developing such a system, researchers rely on input data to train recognition models. In the absence of openly available datasets that meet the model requirements, researchers face a hard and time-consuming process to decide which sensing device to use or how much data needs to be collected. In this paper, we explore the effect of the amount of training data on the performance (i.e., classification accuracy and activity-wise F1-scores) of a CNN model by performing an incremental cross-subject evaluation using data collected from a consumer smartphone and smartwatch. Systematically studying the incremental inclusion of subject data from a set of 22 training subjects, the results show that the model’s performance initially improves significantly with each addition, yet this improvement slows down the larger the number of included subjects. We compare the performance of models based on smartphone and smartwatch data. The latter option is significantly better with smaller sizes of training data, while the former outperforms with larger amounts of training data. In addition, gait-related activities show significantly better results with smartphone-collected data, while non-gait-related activities, such as standing up or sitting down, were better recognized with smartwatch-collected data.
Matey-Sanz, Miguel; Torres-Sospedra, Joaquín; Moreira, Adriano
Temporal Stability on Human Activity Recognition based on Wi-Fi CSI Proceedings Article
In: 2023 IEEE 13th International Conference on Indoor Positioning and Indoor Navigation (IPIN), pp. 1-6, IEEE, 2023, ISBN: 979-8-3503-2012-1.
Abstract | Links | BibTeX | Tags: activity recognition, CSI, machine learning
@inproceedings{Matey2023a,
title = {Temporal Stability on Human Activity Recognition based on Wi-Fi CSI},
author = {Miguel Matey-Sanz and Joaquín Torres-Sospedra and Adriano Moreira},
doi = {https://doi.org/10.1109/IPIN57070.2023.10332214},
isbn = {979-8-3503-2012-1},
year = {2023},
date = {2023-09-25},
booktitle = {2023 IEEE 13th International Conference on Indoor Positioning and Indoor Navigation (IPIN)},
pages = {1-6},
publisher = {IEEE},
abstract = {Over the last years, numerous studies have emerged using Wi-Fi channel state information, enabling device-free (passive) sensing for applications such as motion detection, indoor positioning or human activity recognition. More recently, the development framework for the low-cost ESP32 microcontrollers has added support for obtaining channel state information data. In this work, we collected channel state information data for human activity recognition, where activities are relatively localized with respect to the Wi-Fi infrastructure. The data was collected in several runs, duly spaced in time, and a convolutional neural network model was used for the classification of activities. Classification performance evaluation showed a clear degradation when a model evaluated with data collected 10 minutes after the data used for training showed a 52% relative loss in the accuracy of the classification.},
keywords = {activity recognition, CSI, machine learning},
pubstate = {published},
tppubtype = {inproceedings}
}
Over the last years, numerous studies have emerged using Wi-Fi channel state information, enabling device-free (passive) sensing for applications such as motion detection, indoor positioning or human activity recognition. More recently, the development framework for the low-cost ESP32 microcontrollers has added support for obtaining channel state information data. In this work, we collected channel state information data for human activity recognition, where activities are relatively localized with respect to the Wi-Fi infrastructure. The data was collected in several runs, duly spaced in time, and a convolutional neural network model was used for the classification of activities. Classification performance evaluation showed a clear degradation when a model evaluated with data collected 10 minutes after the data used for training showed a 52% relative loss in the accuracy of the classification.