An Optimised Algorithm for Accurate Steps Counting From Smart-Phone Accelerometry.

Area of Science:

• Biomedical Engineering
• Mobile Health Technology
• Wearable Sensors

Background:

• Smartphone accelerometers offer potential for widespread fitness and health monitoring.
• Accurate step counting is hindered by the variable ways smartphones are carried.
• Existing algorithms require optimization for diverse usage scenarios.

Purpose of the Study:

• To refine and optimize the windowed peak detection algorithm for smartphone-based step counting.
• To improve the accuracy and efficiency of step estimation across various phone carrying positions.
• To develop a novel ground truth device for precise step data collection.

Main Methods:

• Utilized data from three volunteers carrying smartphones in six distinct positions.
• Explored and optimized parameters of the windowed peak detection algorithm.
• Developed and employed a novel device for ground truth step detection.
• Implemented and validated the algorithm on the Android OS as an open-source project.

Main Results:

• The optimized algorithm achieved an average accuracy of 95% across the collected dataset.
• Validation on a separate real-time dataset confirmed the algorithm's robust performance.
• The novel ground truth device facilitated simplified and accurate data analysis.

Conclusions:

• The optimized windowed peak detection algorithm provides accurate and efficient smartphone-based step counting.
• The open-source implementation facilitates broader adoption in mobile health applications.
• This work advances the reliability of step counting for fitness and health monitoring using mobile devices.