Performance Comparison of Discrete Transforms on Speech Compressed Sensing

This work explores the application of compressed sensing technique on data sampling and acquisition of speech signals. Our objective is to provide the use of compressed sensing for speech signals which investigates the possibility of reliable reconstruction of a speech signal from very few sample and also to compare the performance of DFT and DCT on speech signal in compressed sensing. Since DCT representation produces comparatively less number of coefficients, it enables a better sparse representation of speech signal than DFT representation. The same hearing quality of speech signal is recovered by DCT using less number of measured coefficients than required by DFT. Since the coefficients of DCT are real, the reconstruction algorithm of DCT runs comparatively faster than DFT and also requires less time for recovery.