An approximate compressor for wearable biomedical healthcare monitoring systems

Technology advancements as well as the Internet-of-Things paradigm enable the design of wearable personal healthcare monitoring systems. Ultra-low-power design is a challenging area for these battery-operated wearable devices, where the energy supply is limited and hardware resources are scarce. Some biomedical applications tolerate small errors in the values of the biosignal or small degradation in the quality, which can be exploited to reduce the energy requirements. This paper presents an approximate compression technique for biosignals in a wearable healthcare monitoring system. It takes advantage of error tolerance in biosignals and finds the shortest code to compress the data while keeping the error in an acceptable range. Our approximate compressor does not demand any hardware modification and thus can be used in existing wearable devices. The proposed approach for reducing the size of the Huffman table can save 1 MBit storage, on average. It also makes our approximate compressor suitable for runtime adaptation, i.e. creating a new Huffman table based on updated values. Compared to state-of-the-art, our experimental results show up to 60% reduction in data size that is to be transmitted via radio. As wireless communication contributes significantly to the total energy consumption of wearable devices, this improvement can increase the battery lifetime of our healthcare monitoring prototype from 7 days to 10 days.