A Sensor Node Lossless Compression Algorithm based on Linear Fitting Residuals Coding

Abstract

According to the theory of linear regression model, this paper designed a sensor data lossless compression algorithm. The algorithm calculates the sensor data's fitting values and fitting residuals, which are input to a content-based entropy coder to perform compression. The algorithm achieves lossless transform by rounding operation, and realizes positive sequence decoding by prediction fitting. The efficient entropy coding is realized by calculating the mean bit number of input data. Compared with the typical lossless compression algorithms, the proposed algorithm indicated better compression ratios with a small computational overhead.