Improving human activity recognition and its application in early stroke diagnosis

Area of Science:

• Biomedical Engineering
• Computer Science
• Artificial Intelligence

Background:

• Stroke poses significant global health and economic burdens.
• Early diagnosis is crucial for effective stroke management and rehabilitation.
• Human Activity Recognition (HAR) is vital for developing automated stroke detection tools.

Purpose of the Study:

• To develop an efficient Human Activity Recognition (HAR) system for early stroke detection.
• To create a method capable of discriminating between normal resting states and stroke-induced paralysis.
• To enhance existing HAR techniques for improved stroke diagnosis.

Main Methods:

• An extended Genetic Fuzzy Finite State Machine (GFFSM) approach was utilized.
• A novel hybrid feature selection (FS) algorithm was developed, combining Principal Component Analysis (PCA) and a voting scheme.
• Cross-validation results were integrated to refine feature selection.

Main Results:

• The proposed HAR approach demonstrated high performance in distinguishing activities.
• The system successfully identified patterns indicative of stroke-related paralysis.
• Experimental results confirmed the efficacy of the GFFSM extension and hybrid FS algorithm.

Conclusions:

• The developed HAR system is a well-performing tool for early stroke detection.
• The proposed method can be successfully integrated into portable devices for real-time monitoring.
• This research contributes to advancing automated diagnostic tools for neurological conditions.