Smartwatch Accelerometer Step Counting That Rejects False Positives During Non-Walking Wrist Movement.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Health Informatics

Background:

• Wrist-worn devices are crucial for health management and disease prevention.
• Inaccurate step counting from false positives during non-walking wrist movements can lead to flawed health assessments and interventions.

Purpose of the Study:

• To develop a real-time adaptive multi-stage step counting algorithm for smartwatches.
• To integrate non-walking movement detection to eliminate false positive step counts.

Main Methods:

• A smartwatch with a 3-axis accelerometer was used.
• Sixty-seven subjects performed walking, running, and eight distinct non-gait wrist movement trials.
• A novel algorithm was developed to differentiate between walking and non-walking movements.

Main Results:

• The proposed algorithm achieved 93.58% accuracy in step counting, significantly outperforming a standard peak detection framework (10.09%).
• The algorithm demonstrated superior accuracy and efficiency in non-walking detection compared to four baseline models.
• Inference time was only 8.9% of a single OCSVM, indicating high efficiency.

Conclusions:

• Rejecting false positive step counts from non-walking movements is critical for accurate wrist-worn step counting.
• The proposed algorithm offers potential for more precise step estimation in real-world scenarios.
• This technology can enhance aerobic exercise assessment and aid in preventing sedentary diseases.