An accelerometer lossless compression algorithm and energy analysis for IoT devices

The Internet of Things promises to enable numerous future applications spanning many domains, including health care, and is comprised of devices that are constrained in terms of computational and energy resources. A specific health care application is to ascertain patients' activity of daily living while at home using accelerometer data from non-invasive wearables. It is often necessary to store this data on the device to be retrieved later for analysis. However, the devices typically store far more data than can be transmitted with commonly used low power radios. To mitigate the problem, this paper proposes an energy efficient, lossless compression algorithm that uses an offline frequency distribution to create a symbol-code lookup table. Using an extensive set of data from a previous study, an analysis of the entropy of activities of daily living accelerometer data is presented. The compression algorithm is compared against this estimated entropy. Energy being critical for IoT devices, the trade-off between energy cost for compression versus energy saved during transmission is also analysed.