Channel Vocoder 2v2u4g

CHANNEL VOCODER

Submitted by Srijith. V Paul James Mithun Mathew

CHANNEL VOCODER The oldest form of speech coding device is the channel vocoder which was invented by Dudley. The channel vocoder incorporates more of speech model into the analysis and synthesis configuration and for simplicity a number of approximations are introduced into the time dependant fourier analysis and synthesis. Each channel analysis can be thought of as a band filter with center frequency, w

CHANNEL VOCODER- ANALYZER

The above block diagram is than of a channel vocoder analyzer. The full wave rectifier and the LPF act as an appropriate envelope detector. Such a system is the basic component of a channel vocoder. The analyzer consists of a bank of such channels with analysis frequencies distributed across the speech band of interest. But speech signals cannot be represented by the amplitude spectrum alone and the phase derivative consists of information about the excitation. It is known that if the phase derivative signals are arbitrary set to 0, the resulting signal will be a monotone. In order to represent the proper excitation of the signal the analyzer has an additional component for determining the mode of excitation, ie; voiced or unvoiced and if voiced, the fundamental frequency of the speech signal. The resulting excitation information along with the amplitude channels signals, form the representation of the speech signal. These parameters are sampled and quantized for storage and transmission in a digital system.

CHANNEL VOCODER- SYNTHESIZER

The channel signals controls the amplitude of contribution of a particular channel, while the excitation signals control the detailed structure of the output of a given channel. The voiced or unvoiced signal simply serves to select an appropriate excitation generator.ie; random noise for unvoiced speech and periodic pulse generator for voiced speech, while the fundamental frequency of the pulse generator is being controlled by the pitch signal. Thus the composite output spectrum is built up of individual segments in which the amplitude within a given frequency band is roughly constant. In fact the amplitude in a particular band retains the frequency selective shaping properties of the band filter used for synthesis. If the excitation is voiced the output is composed of contiguous bands in which the final spectral structure is characteristic of the periodicity while if the excitation is unvoiced, the spectrum varies continuously over every band. The resultant speech signal appears to be highly reverberant in nature because of the complete lack of control over the merging together of adjacent bands.