Speech coding and decoding method using high-order linear prediction coefficient grouping vector quantization

Abstract

The invention discloses a speech coding and decoding method using high-order linear prediction coefficient grouping vector quantization. The method can be applied to a low-bit-rate and high-quality speech codec and can be used to reduce the codebook size and computation load required for high-order line spectrum frequency vector quantization

Description

technical field [0001] The invention relates to a voice encoding and decoding method using high-order linear prediction coefficient grouping vector quantization, and belongs to the technical field of voice encoding. Background technique [0002] The basic voice communication process is shown in the attached manual. figure 1 As shown, the requirement for the voice codec is: under the premise of ensuring the voice quality, the lower the bit rate of the coded stream, the better. Codecs that use vector quantization techniques, linear predictive analysis techniques, and data interpolation techniques can well meet this requirement. In the speech codec based on the above technology, the encoding and decoding process of the linear prediction coefficient is as follows: at the encoding end, use the Levinson-Durbin algorithm to solve the linear prediction coefficient, and convert it into the corresponding line spectrum frequency parameter, and use the storage Vector quantization is p...

AI Technical Summary

Problems solved by technology

However, there are still two problems: first, directly calculating the high-order line spectrum frequency parameters from the high-order linear prediction coefficients requires a large amount of computation, which is proportional to the square of the number of linear prediction coefficients; second, for T Vector, T codebooks need to be designed separately, so the total amount of codebooks required for high-order line spectrum frequency parameters is still large

However, as the order of linear prediction increases, the existing speech encoders use the algebraic equation root method to solve the line spectrum frequency parameter lsf(i) from the linear prediction coefficient a(i), and calculate the line spectrum frequency parameter from the linear prediction coefficient. The amount of calculation is proportional to the square of the number of linear prediction coefficients; at the same time, even if the split vector quantization technology is used, the codebook size required for high-order line spectrum frequency vector quantization is also large, so high-order linear prediction analysis in low code application in high-rate, high-quality speech coders is very difficult

Images