Registration and Analysis of Acceleration Data to Recognize Physical Activity.

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Wearable Technology

Background:

• Wearable sensors, particularly smartphones, offer a non-invasive method for monitoring physical activity.
• Accurate physical activity recognition is crucial for health monitoring, fitness tracking, and rehabilitation.

Purpose of the Study:

• To evaluate the efficacy of smartphone-recorded acceleration signals in distinguishing between various physical activities.
• To assess both user-independent and user-dependent classification accuracies.

Main Methods:

• Experiments involved recording acceleration signals from a hip-worn smartphone during five distinct physical activities: running, standing, ascending stairs, descending stairs, and walking.
• Feature extraction was performed using statistical parameters of the acceleration signals.
• Quadratic Discriminant Analysis (QDA) was employed for activity classification.

Main Results:

• User-independent classification accuracy for five physical activities reached 83%.
• User-dependent classification accuracy demonstrated a high range, from 90% to 95%.
• The study confirmed that acceleration data from a hip-worn device effectively differentiates between multiple physical activities.

Conclusions:

• Smartphone-based acceleration sensing is a viable and effective method for recognizing diverse physical activities.
• The findings support the use of wearable accelerometers for objective physical activity assessment.
• High classification accuracies suggest practical applications in health and wellness monitoring.