ABSTRACT

The speech signal, or speech wave, can be changed into a processible object by converting it into an electrical signal using a microphone. The electrical signal is usually transformed from an analog into a digital signal prior to almost all speech processing for two reasons. First, digital techniques facilitate highly sophisticated signal processing which cannot otherwise be realized by analog techniques. Second, digital processing is far more reliable and can be accomplished by using a compact circuit. Analog-to-digital conversion, commonly referred to as digitization, consists of the sampling, quantizing, and coding processes. The short-time analysis interval multiplied by a window function and extracted from the speech wave is called a frame. Sound spectrogram analysis is a method for plotting the time function of the speech spectrum using density plots. In speech analysis-synthesis systems, it is necessary to extract source parameters in parallel with spectral envelope parameter extraction.