Tunable-Q wavelet transform based optimal compression of cardiac sound signals

In this paper, we introduce a new approach for compression of cardiac sound signals using tunable-Q wavelet transform (TQWT) for efficient telemetry based monitoring and diagnosis of heart disorders and data archiving. In the proposed method, the cardiac sound signals have been compressed using TQWT, linear quantization, Huffman and run length coding (RLC) techniques. To begin with, the cardiac sound signals have been decomposed using TQWT. Then, a dynamic threshold has been applied on the obtained wavelet coefficients to achieve distortion error in acceptable range. The wavelet coefficients above the threshold and the corresponding binary significant map have been compressed by steps involving zero removal, linear quantization/RLC and Huffman coding. Optimal values of the compression parameters have been found using genetic algorithm (GA) with a subset of dataset. The performance of these optimized values of compression parameters have been evaluated using a test set. The proposed compression method has provided significant compression performance with lower distortion for various clinical cases as comprised in the publicly available dataset. Moreover, the obtained results have been found comparatively better than that of an existing wavelet transform (WT) based method due to the properties of TQWT and the resulting increased number of compression parameters for optimization.