Smart-Sleeve: A Wearable Textile Pressure Sensor Array for Human Activity Recognition.

Area of Science:

• Wearable Technology
• Human-Computer Interaction
• Biomedical Engineering

Background:

• Human activity recognition is crucial for various applications.
• Sensing human interaction with the environment is key to understanding activities.
• Existing methods may lack unobtrusiveness or comprehensive sensing capabilities.

Purpose of the Study:

• To propose and evaluate a Smart-Sleeve for sensing human contact and recognizing activities.
• To develop a comprehensive dataset for human activity recognition.
• To assess the performance of classical and deep learning models using the Smart-Sleeve data.

Main Methods:

• A Smart-Sleeve composed of a soft, stretchable textile pressure sensor matrix was developed.
• A dataset of 18 activities performed by 14 subjects (10 repetitions each) was created.
• Performance was evaluated using six classical machine learning classifiers and a convolutional neural network, employing a novel normalization approach and statistical, geometric, and symmetry features.

Main Results:

• The Smart-Sleeve demonstrated suitability for human activity recognition.
• High average accuracies were achieved: 82.02% (SVM) and 82.30% (CNN) in 10-fold cross-validation.
• Leave-one-subject-out validation yielded accuracies of 72.66% (SVM) and 74.84% (CNN).

Conclusions:

• The proposed Smart-Sleeve effectively captures human-environment contact for activity recognition.
• The developed data processing methods and wearable sensor system are robust.
• This technology shows significant promise for practical human activity recognition applications.