Lossless compression of electromyographic signal

For long-term electromyographic (EMG) signal acquisition, data compression can be used to reduce the original data to a smaller size. This reduction decreases the bandwidth and energy used during transmission. Storing the compressed data on storage media also uses less space. This study aims to determine suitable lossless compression algorithms for EMG signal on small embedded devices. These devices have much less processing power than laboratory-sized devices thus a proper tool has to be carefully chosen. Various algorithms are studied with a set of simulated and real EMG signals. The performances of the algorithms are measured and compared. The memory requirement of the algorithms are also recorded for future reference. The result shows that the LPC based algorithms performs very well with EMG signal. The average compression ratio of these tools are 2.44. While BWT and LZMA algorithms are comparable to LPC in term of CR, these tools are very complex and have a much higher resource requirement. In this study, FLAC outperforms other compression tools with average CR of 2.61. This tool is fast and consumes slightly more resource than a much simpler scheme like SHN. For the standard LPC modeling, choosing an order of the model greater than 4 produces insignificant benefit. The simpler SHN scheme performs 4.2% better than the best LPC.