Speech Coding-simple understanding

Speech Encoding is divided into three categories:

1. waveform Encoding

PCM and ADPCM belong to waveform encoding. This encoding method is used to directly quantify each sample point of the waveform, or to compress the correlation between waveforms to remove redundancy and ensure better speech quality, however, the encoding speed is high and the compression speed is small.

2. Parameter Encoding

Such as LPC encoding, formant encoding, and Vocoder encoding are parameter encoding. This encoding method is used to model the speech signal, extract the acoustic parameters that represent the speech of this Section, encode the acoustic parameters that represent the model, decode the model parameters at the decoding end, and reconstruct the model based on the reconstruction model, returns the voice waveform. Obviously, this encoding method is characterized by high compression and low speed requirements. However, the speech quality is generally not highly waveform encoding, which mainly depends on the accuracy of the model decomposition and reconstruction.

3. Mixed Encoding

CELP belongs to the class encoding type. I am not very familiar with this encoding method. Since it is a mixture, it is a compromise between waveform encoding and parameter encoding. On the basis of low-to-medium-rates, we use complicated algorithms to obtain high-quality speech as much as possible.