Compressed Sensing of Skin Conductance Level for IoT-based wearable sensors

In the context of physiological monitoring applications, wearable sensor platforms usually have to sustain long-term acquisitions, necessary to assess correctly the individual health status, but also to compensate for uncertainty and noise effects, due to the uncontrolled conditions in which data are collected. Therefore, smart measurement systems based on the Internet of Things paradigm require approaches able to optimize the on-board resources, either in terms of computational effort and power consumption. With a view to ensuring such an optimization, this paper presents a Compressed Sensing (CS) method for the reconstruction of the Skin Conductance Level, a signal able to encode individual traits of observed subjects, useful also in bio-metric applications. The proposed CS method employs a deterministic measurement matrix and it exploits the sparsity in the domain defined by the Discrete Cosine Transform. The Percentage of Root-mean-squared Difference, chosen as figure of merit, shows very low values for a Compression Ratio as high as 12, thus proving the excellent performance of the proposed method.