Multi-speed audio encoding method

Abstract

The embodiment of the invention discloses a multi-rate voice frequency encoding method. The method comprises the following steps: computing frequency spectrum of a perceptual weighting filter and a first ratio of the frequency spectrum of two previous layers of synthetic voice which are firstly encoded and then decoded by an input signal for each adjacent lattice point; and encoding index values of the corresponding lattice point of the first ratio into a code stream according to a descending order of the first ratio. The method can help improve the quality of output voice or sound signals when lacking bits during transform domain encoding.

Description

technical field [0001] The present invention relates to encoding technology, in particular to a multi-rate voice and audio encoding method. Background technique [0002] In the current multi-rate speech and audio coding, when the input signal is more in line with the music characteristics or the code rate is high (for example, in AMR-WB+, G.729.1 and G.VBR), transform domain coding is often used, that is The time-domain signal is transformed into the frequency domain by a transform method, such as Modified Discrete Cosine Transform (MDCT) or Fast Fourier Transform (FFT), etc. When quantizing the parameters of the transform domain coding, the lattice vector quantization technology is mostly used. figure 1 It is a flowchart of a lattice vector quantization method in the prior art. Such as figure 1 As shown, in general, lattice vector quantization includes the following steps: [0003] Step 101, find the corresponding grid point for the frequency domain signal according to ...

AI Technical Summary

Problems solved by technology

Therefore, when the number of bits is insufficient, only part of the code stream can be decoded at the decoding end, and because the parameters of the transform domain encoding at the encoding end are written in the code stream in the order of low frequency to high frequency, only a small amount can be recovered at this time. low-frequency signal, resulting in that when the bit rate increases, the quality of the output voice or audio signal does not improve significantly

In addition, when calculating the index value of the grid point, the importance of the grid point is determined according to the size of the spectral energy. When the number of bits is not enough, the elements of the grid point with smaller spectral energy are set to 0; but the part with larger spectral energy It is not necessarily an important component. Such a judgment criterion may set the elements of the grid points corresponding to some important components to 0, affecting the quality of the output signal.

In addition, according to the principle of masking effect, it can be seen that whether a signal is masked or not depends entirely on the size of the spectrum energy, to a certain extent, it also depends on the difference between the masked signal and the masked signal, and there is no such thing in the above coding method consider this difference

Images