Work

Towards improved human arm movement analysis: Advanced feature engineering and model optimization

Year: 2025

Type: article

Abstract: Human arm movement analysis plays a pivotal role in rehabilitation, sports science, and human-computer interaction. This study focuses on the development and evaluation of advanced feature extraction, reduction, and selection techniques for classifying time series data collected from human arm movements. Targeting key motions such as elbow extension/flexion, shoulder extension/flexion, and wrist pronation/supination, data were captured using accelerometers and gyroscopes. These sensors yielded comprehensive motion data, which underwent preprocessing including noise reduction and segmentation via a sliding window approach. A combination of time-domain and frequency-domain feature extraction was performed, yielding features such as statistical measures, and frequency components. Embedded methods were applied for feature selection, including Random Forest, Least Absolute Shrinkage and Selection Operator, and Elastic-Net. Our combined feature selection model achieved an accuracy of over 96% on the reduced feature set, decreasing the number of features from 720 to 46, and highlighting the critical importance of rotational data from gyroscopes. The approach effectively enhanced classification accuracy while maintaining computational efficiency. The comprehensive framework developed in this study offers robust methodologies for feature extraction, reduction, and selection, demonstrating significant improvements in the classification of human arm activities. The findings have practical implications in developing personalized rehabilitation programs, optimizing athletic performance, and advancing gesture-based control systems. Future work will focus on integrating more sensors, as electromyography, to enhance robustness and applicability of the proposed methods. Additionally, the development of real-time monitoring systems for rehabilitation exercises, leveraging wearable devices, could significantly improve therapy outcomes by providing immediate feedback and enabling continuous tracking.

Source: Engineering Applications of Artificial Intelligence

DOI: 10.1016/j.engappai.2025.111194

Authors: Yassine Benachour, Moez Rehman, Farid Flitti

Institution: Higher Colleges of Technology

Language: English

Topic: Context-Aware Activity Recognition Systems

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

SDG: Good health and well-being

Open Access status: hybrid

Estimated APC paid: $3,200