Speech Signal Compression and Reconstruction Using Inverse Technique

Speech compression is a process of compressing speech signal to reduce its size for transfer. This paper proposed a new technique to compress the speech signal. This technique is called the decimation process. It is opposite of interpolation. This process reduces the sampling rate and thus save time, storage capacity, and cost. Decimation contains two stages, processes of lowpass filtering followed by downsampling. The benefit of using a filter is to avoid aliasing effect. The reconstruction of the original speech signal can be performed using inverse interpolation techniques such as maximum entropy and regularization theory. Finally, we assess the quality of the reconstructed signal using quality metrics such as signal-to-noise ratio (SNR), signal to noise ratio segmental (SNRseg), spectral distortion (SD) and log-likelihood ratio (LLR).