A 1.2kb/s low-rate speech codec method based on mixed excitation linear prediction melp

Abstract

The invention relates to a 1.2kb / s low-rate speech encoding and decoding method based on mixed excitation linear prediction MELP, comprising the following steps: an encoding end frames a speech signal with 30ms as the sub frame length, groups two adjacent sub frames into a super frame, and carries out multi-frame joint quantization encoding on extracted speech feature parameters LSF, Pitch, VP, Fsmag and G; the encoding end uses the remaining bits in the frame structure to carry out error control encoding on important speech feature parameters, and finally, a binary bit stream is formed and transmitted; and a decoding end parses out the quantitative index values of the speech feature parameters from the received bit stream, works out the initial values of the speech feature parameters through quantitative indexing, then, carries out speech feature parameter integrity reconstruction, generates an excitation signal based on the reconstructed speech feature parameters, and obtains a synthetic speech signal through adaptive spectral enhancement, a synthesis filter, gain control, and distribute pulse filtering. By adopting the method, the speech encoding rate can be effectively reduced. The speech synthesized by a receiving end has high clarity and intelligibility.

Description

technical field [0001] The invention belongs to the field of voice communication, and in particular relates to a MELP-based low-rate voice codec, which is widely used in voice application services such as secure communication, satellite mobile communication, deep-sea communication and voice mailbox. Background technique [0002] Voice is the main carrier for human beings to transmit information to each other. It is the most direct, convenient and effective way of communication in modern communication, and it is also the main means of human-computer interaction in the future. With the development of communication technology, the proportion of non-voice information such as images and data in information transmission is increasing, but the effective transmission of voice information is still one of the necessary functions of many communication systems. [0003] Although the introduction of optical fiber transmission technology in recent years has provided huge transmission capa...

AI Technical Summary

Problems solved by technology

NATO's 1.2kbps vocoder uses three 20ms subframes to form a superframe for joint quantization. Due to the large number of subframes, there are many parameters extracted from each subframe, and the encoding output of each superframe is only 72bit. Therefore, the sound The coder adopts the vector quantization coding technology for the pitch period, the line spectrum for the frequency, the residual harmonic amplitude value, the gain and the band-pass voiced sound intensity, which leads to the increase of the coding complexity and the memory capacity required for storing the vector quantization codebook, and the acoustic Encoder implementation costs increase

Images