Lossless compression of vibration signals on an embedded device using a TDE based predictor

Abstract

A lossless compression scheme for the data acquired from three-axial microelectromechanical (MEMS) accelerometer is presented. Time delay estimation (TDE) was applied in conjunction with differential pulse code modulation (DPCM) as the preprocessor to entropy coding, to perform lossless compression on an embedded sensor device with limited memory and speed in real time. The essence of this method is to check if the signal exhibits certain level of periodicity, and code the differences between samples one signal period apart if it does. Limited choice of feasible mathematical operations was an important constraint implied by the architecture of the embedded sensor device. Algorithm execution time was improved by programming in assembler and avoiding unnecessary operations. Experiments have confirmed that considerable compression ratio gains can be achieved if the signal is quasiperiodic.