Short time spectrum analysis-synthesis of speech

This thesis presents an extensive study of the short time spectrum analysis-synthesis of speech. This type of speech processing attempts to synthesize speech without the short time Fourier transform phase information. Several existing analysis-synthesis methods are presented. In addition, five other methods are attempted experimentally. The results of these experiments are explained, identifying some very important properties of the short time spectrum analysis-synthesis. One of the methods was shown to synthesize a high fidelity speech. Finally, a comparison is made between this method and other known methods which are based on short time Fourier transform approach