Adaptive lossless compression in wireless body sensor networks

In most wireless body sensor network (BSN) applications, the vast majority of the total energy is consumed by the wireless transmission of sensed data. Transmitting one bit using a typical wireless communication system can consume as much energy as 1000 cycles of an embedded processor. Reducing this transmission energy – even at the expense of increasing another component’s energy – is essential to meeting the battery life and form factor (i.e. small battery) requirements of many BSN applications. While improved wireless communication and networking techniques can help do just that, simply compressing the sensed data to reduce the number of transmitted bits can provide significant savings. However, BSN platforms and applications impose many constraints on compression techniques, including fidelity (focus on lossless techniques, as required for many medical BSN applications), programmability (enable ease of code development and deployment), adaptability (achieve high compression ratio regardless of location, subject, activity, etc.), and implementability (require low processing and memory resources). This paper analyzes variations of two known real-time lossless compression algorithms, Huffman encoding and delta encoding, within the context of these BSN constraints. Experimental results on a multi-node accelerometer-based BSN show the strengths and weaknesses of each algorithm and ultimately reveal the superiority of dynamic delta encoding for BSNs, including an average 35% energy savings across a range of activities, sensor locations, and sensor axes.