Abstract
Speech compression is the process of compressing a speech signal to reduce its size for easy transmission. This paper investigates two techniques to compress the speech signal and reconstruct it. The first technique is based on a decimation process. This process reduces the sampling rate and consequently saves time, storage capacity, and cost. The decimation process comprises a low-pass filtering stage followed by a down-sampling stage. The recovery of the original speech signal from the decimation-based compressed signal can be accomplished using the inverse interpolation techniques, namely, the maximum entropy and the regularization methods. The second technique is based on the concept of compressed sensing (CS). CS has a very vital role in signal compression and reconstruction because of its ability to sample the signal at a rate smaller than the Nyquist rate. It depends on non-adaptive linear projections that save the structure of the signal and the reconstruction of the original signal is performed by solving a linear optimization problem. Finally, the quality of the recovered signal is assessed using signal-to-noise ratio (SNR), spectral distortion (SD), and correlation coefficient (c (r) ).