Accelerometer-Based Automated Counting of Ten Exercises without Exercise-Specific Training or Tuning

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Wearable Sensors

Background:

• Wearable sensors are crucial for tracking physical activity in clinical research and personal motivation.
• Existing exercise counting models often require specific training for each movement, limiting their application in physical therapy where unique motions occur.

Purpose of the Study:

• To develop an automatic repetition counting system for physical therapy exercises.
• To create a flexible system capable of measuring diverse, repetitive movements without prior motion-specific training.

Main Methods:

• Utilized smartphone accelerometers attached to participants to record repetitive motions during 10 different exercises.
• Applied three analysis techniques: threshold crossing, threshold crossing with low-pass filtering, and Fourier transform to count repetitions.
• Evaluated system performance across 18 participants performing exercises like arm circles, squats, and push-ups.

Main Results:

• The automatic counting system demonstrated effectiveness in tracking exercises like arm circles and push-ups.
• Less periodic and irregular motions, such as upper trunk rotations, presented greater tracking challenges.
• Threshold crossing with low-pass filtering emerged as the most effective method for repetition counting.

Conclusions:

• The developed automatic counting system can track exercise repetitions effectively without requiring pre-training for specific activities.
• This technology offers a promising solution for monitoring physical therapy progress and promoting adherence to exercise regimens.