Lightweight Compressed Sensing (CS) and Partial DCT Based Compression Schemes for Energy-Efficient Wearable PPG Monitoring Devices

Most wearable medical devices are designed to continuously acquire photoplethysmography (PPG) for measuring vital signs and transmitting acquired PPG data wirelessly to edge-computing device or cloud-computing server. These devices are constrained with limited battery power and data-rate capacity. Therefore, in this paper, we present a lightweight effective data-reduction method by investigating the performance of compressed sensing (CS)-based and and partial discrete cosine transform (DCT)-based compression methods with major objectives of achieving higher compression ratio (CR) with minimal waveform distortion with low reconstruction time. By using both normal and abnormal PPG signals, the performance of the CS-based and DCT-based compression methods is evaluated in terms of CR, global and local distortion measures and processing time. Evaluation results showed that CR values of the partial-DCT based method are 3 times higher (CR ranging from 7.50 to 9.38) without distorting fiducial points and shapes of the PPG signal (percentage root-mean-square difference (PRD) ranging from 1% to 2%) as compared to the CS-based data method (CR from 2.50 to 3.13 for PRD from 2% of 4%). The higher data reduction with acceptable level of reconstruction quality demonstrates that the partial DCT-based method can lead to provide better overall energy consumption reduction solution for resource-constrained wearable devices.