A Signal Quality Assessment Metrics for Vibration-based Human Sensing Data Acquisition

Abstract

Sensing signal quality affects signal processing efficiency, feature extraction, and learning accuracy. An efficient and accurate assessment of sensing system signal quality is essential for 1) large-scale cyber-physical system deployment and 2) datasets sharing and comparison. In this paper, we present a signal quality assessment -- S-score -- for vibration-based human sensing applications from two aspects -- the hardware implementation and the deployment structure. The 1) signal-to-noise ratio and 2) the signal frequency response consistency over 2.1) sensing hardware, and 2.2) deployment structure are essential factors for structural vibration sensing signal evaluation. The S-score metrics combines these factors to a value between 0 and 1 with application-oriented weights. We compared the proposed metrics to two baselines, and our metrics achieved the highest correlation to the system performance, which is the indicator of the data quality.