Wavelet based vector quantization with treecode vectors for EMG Signal compression

Abstract

This paper presents a wavelet-based vector quantization technique using DCCR (Distortion constrained codebook replenishment) mechanism for compression of Electromyogram (EMG) signals. Wavelet coefficients, obtained from EMG signal samples, are arranged to form tree vectors (TVs), where each vector has a hierarchical tree structure. Vector quantization is then applied for encoding to TVs, which uses a pre-calculated codebook. Codebook is created using codebook training algorithm and is updated dynamically using SPIHT coding strategy. Signal is decoded using a copy of the same codebook available with encoder. Tests were performed on EMG records obtained from PGI, Chandigarh. A good quality of reconstructed signal and sufficient compression is achieved. An average Compression Ratio (CR) of 20.64:1 at percentage root mean square difference (PRD) of 6.12% is obtained by this technique.