Microcontroller Implementation of Low-power Compression for Wearable Biosignal Transmitter

This paper presents a low-power algorithm for the compression of biosignals (ECG, EMG, gait pattern). Sample decimation is guided by the second derivative of the signal as a metric for signal activity. Here, we describe the implementation of the algorithm on a general purpose microcontroller of the transmitter node of a body-area network. The algorithm is optimized for low computational complexity and consists of 180 controller instructions. It incorporates feedback to achieve a targeted compression factor (CF). Simulated and measured results with different biosignals confirm that the code achieves a typical CF around 10, needs 137 controller cycles per input sample and consumes 507 nJ per sample on the prototype hardware.