Estimation of downsampling influence on accelerometer signal quality

Abstract

Downsampling plays a crucial role in the field of accelerometer data analysis, especially in applications that require optimizing computing resources while ensuring minimal loss of signal integrity. This study examines the influence of downsampling on accelerometer signal quality, critical for optimizing activity recognition and health monitoring applications. Utilizing the Sisfall dataset, which captures diverse physical activities through triaxial accelerometers, the research evaluates the influence of downsampling on signal quality using metrics such as Mean Squared Error (MSE), energy, and Cumulative Explained Variance (CEV). The findings reveal that a specific downsampling factor, effectively balances data volume reduction with signal integrity preservation. This optimal factor enhances computational efficiency without significantly compromising signal quality, pivotal for the accurate analysis and recognition of physical activities. The study underscores the importance of selecting an appropriate downsampling rate to maintain the fidelity of accelerometer data in wearable technologies and health monitoring systems, providing a guideline for signal processing optimizations in such applications.