Physical activity group classification algorithm using triaxial acceleration and heart rate.

Area of Science:

• Biomedical Engineering
• Sports Science
• Wearable Technology

Background:

• Accurate physical activity classification is crucial for estimating energy expenditure.
• Existing methods often lack precision, especially when integrating diverse physiological and motion data.
• Developing robust algorithms is key for applications in health monitoring and fitness tracking.

Purpose of the Study:

• To propose a novel physical activity classification algorithm for improved energy expenditure estimation.
• To enhance classification accuracy by integrating triaxial acceleration and heart rate data.
• To validate the algorithm's performance across various physical activities.

Main Methods:

• Developed a decision tree algorithm utilizing three key indices: heart rate reserve (%HRreserve), filtered triaxial acceleration, and the ratio of filtered to unfiltered acceleration.
• Calculated %HRreserve using resting and maximum heart rate, with maximum heart rate determined by the Karvonen Formula.
• Classified activities into five categories: sedentary, household, moderate (excluding locomotive), locomotive, and vigorous.

Main Results:

• The proposed algorithm achieved an average classification accuracy of 91% for 21 distinct activities.
• The integration of both acceleration and heart rate data significantly improved classification precision.
• Optimized classification directly contributed to more accurate energy expenditure estimations.

Conclusions:

• The developed algorithm offers a more accurate method for physical activity classification and energy expenditure estimation.
• Combining physiological (heart rate) and motion (acceleration) data provides a synergistic approach to activity recognition.
• This algorithm has potential applications in personalized health, fitness, and rehabilitation.