Speech coding

Speech coding is the conversion of analog signals into digital signals by coding, also known as digital phonetization. The effect of coding rate reduction is achieved by using human auditory characteristics and signal redundancy existing in the process of human vocalization. As much as possible to achieve a smaller communication capacity while achieving higher voice transmission quality.

There are three main speech coding methods, which are waveform coding, parametric coding and mixed coding.

Waveform coding

Waveform coding is the simplest and earliest used. The main principle is to derive the corresponding digital coding form according to the speech signal waveform, sample the analog speech signal at a certain rate on the time axis, and then quantize the amplitude sample hierarchically and express it with code. It does not make use of any parameters that generate the audio signal and directly transforms the time domain signal into a digital code. The purpose of waveform coding is to make the reconstructed speech waveform as consistent as possible with the waveform shape of the original speech signal. It has the advantages of simple method, easy implementation, strong adaptability and good speech quality. However, due to the simple coding method, it also brings some problems. It has high coding rate and low coding efficiency, usually above 16kbps, and the quality is quite high.

Parametric coding

Parametric coding, also known as vocoder technology, is to establish a speech signal generation model and extract the parameters representing the characteristics of the speech signal to encode, in order to reconstruct the speech signal with the highest possible intelligibility and maintain the semantics of the original speech signal, but not necessarily match the original signal on the waveform. Parametric coding is based on the digital model generated by the speech signal, and then the model parameters of the digital model are calculated, and then the digital model is restored according to these parameters, and then the speech is synthesized. The parameter coding rate is low, which can be as low as 2.4kbps or even below. Since the generated speech signal is restored by the established digital model, the reconstructed speech signal waveform may be quite different from the original speech signal waveform, and the distortion will be relatively large. Moreover, due to the limitations of the speech generation model, increasing the data rate cannot improve the quality of the synthesized speech, which has low naturalness and is sensitive to the noise of the speaking environment. Although the sound quality of parameter coding is relatively low, it has good security and stability. It has been widely used in military. The typical parametric Coding method is Linear Predictive Coding (LPC).

Hybrid coding

Hybrid coding is the organic combination of waveform coding and parametric coding, it breaks through the boundary of waveform coding and parametric coding, and combines the high quality of waveform coding and the low coding rate of parametric coding, enhances the naturalness of reconstructed speech, and improves the quality of speech significantly. The disadvantage is that the coding rate increases accordingly.

Your task

Choose a coding method that interests you and talk about it.

*Share your thoughts and ideas in the comments below.