Bandwidth extension coding and decoding method and apparatus

Abstract

The invention relates to a bandwidth extension coding and decoding method and apparatus. The combination of adaptive multiresolution filtering and adaptive time-frequency grid structure and a plurality of linear prediction encoding high-frequency details generates two key technologies, so that the high-frequency part encoding efficiency of digital audio signals and the sound quality of high-frequency part signals can be obviously improved, the low-frequency part of the digital audio signals can still be encoded by means of the traditional perceptual audio coding (for example DRA), and the encoding technology with high subjective sound quality at low bit rate and medium bit rate is realized. In another aspect, an enhancement tool is added on the basis of a current high-quality perceptual coding algorithm such as the DRA, in this way, the downward compatibility with the traditional perceptual coding DRA or other algorithms can be guaranteed. A digital audio encoding and decoding device based on the bandwidth extension coding and decoding method and apparatus can be applied to the fields of satellite HDTV sound accompaniment processing and high-quality audio broadcasting.

Description

technical field [0001] The present invention relates to digital audio codec technology, more specifically, to a bandwidth extension coding and decoding method and device. Background technique [0002] The stereo typical working bit rate of traditional perceptual audio coding technologies (DRA, AAC, MP3, etc.) is 96-128kbps, and the coding quality has obvious subjective perception distortion when it is below 64kbps / stereo. The typical coding bit rate of FM broadcasting application is 48kbps~64kbps / stereo. At this time, the subjective sound quality of traditional sensory audio coding technology can no longer meet the requirements of FM broadcasting. [0003] To this end, a bandwidth extension (BandWidthExtension, BWE for short) coding technique for digital audio signals is proposed. There are many bandwidth extension coding technologies at present, and their performances are also uneven. The bandwidth extension coding techniques that have been published and used in internati...

AI Technical Summary

Problems solved by technology

[0005] The main disadvantages of MPEGSBR coding technology are: (1) the time-frequency division is relatively fixed

Due to the extremely complex audio signal, this algorithm sometimes cannot well meet the accuracy requirements of signal analysis

(2) The high-frequency details of SBR are directly copied from the low-frequency part or obtained by simple filtering of the low-frequency sub-band. At (very) low code rates, this method can greatly reduce the coding rate of the high-frequency part, but Since the low frequency and high frequency of the audio signal of each channel have only a small probability of similarity, the details of SBR high frequency are relatively rough. Although other technologies are applied to reduce the distortion, it is still difficult to restore the entire high frequency part. get higher quality

[0007] The bandwidth extension coding technology of 3GPP AMR-WB+ has the following problems: (1) Realizing high-frequency coding in the time domain cannot obtain higher frequency resolution, because this method can be considered as only one high-frequency area division; (2) The starting frequency band of the encoding is fixed and can only be Fs / 4. For the sampling frequency of 48khz, the starting frequency of high-frequency encoding is 12khz; (3) the harmonic signal in the high frequency cannot be accurately recovered; (4) the high frequency The envelope restoration of the signal is not accurate enough

[0008] In addition, there are some bandwidth extension coding techniques. The time-frequency transformation unit uses traditional FFT, and then divides the high frequency into several regions in the frequency domain, and encodes the spectral energy of each region, so each frame can only provide a time-resolved rate multiple frequency resolution

This FFT-based high-frequency reconstruction technology has high frequency-domain resolution but too low time-domain resolution. When a rapidly changing audio signal is input, the high-frequency reconstructed signal cannot track the changes of the original audio signal well.

Images