Method for acquiring horizontal azimuth parameter codebook in three-dimensional (3D) audio

Abstract

The invention discloses a method for acquiring a horizontal azimuth parameter codebook in a three-dimensional (3D) audio. The method comprises the following steps of: 1, determining the number n of codewords in a required codebook and the dimension number D of each codeword; 2, dividing a horizontal azimuth parameter in a horizontal azimuth parameter training set into D-dimensional vectors to obtain a training vector set, and forming n D-dimensional temporary codebooks by using the first n vectors; 3, dividing the vectors in the training vector set into n cells; 4, calculating subjective average distortion Dsp, comparing the subjective average distortion Dsp and a preset threshold error epsilon, determining a current temporary codebook is a final codebook if Dsp is smaller than epsilon, otherwise executing a step 5; and 5, reselecting a new temporary codebook, and repeating the steps 3 to 4. When the codebook acquired by the method is applied to horizontal azimuth parameter quantification, the whole subjective quantification performance can be improved.

Description

technical field [0001] The present invention relates to the field of quantized codebooks, and more specifically, to a method for acquiring a horizontal orientation parameter codebook in 3D audio. Background technique [0002] Quantization is an extremely important step in audio compression. Quantization is the reduction of data volume by representing a larger data set with a smaller data set. Smaller data sets are often called codebooks, and the data in the codebooks are called codewords. Larger datasets are collections of values ​​to quantify. The data has been quantized, and the original input data is replaced by codewords. It can be seen that the quality of a codebook largely determines the performance of a quantizer. [0003] The operation process of codebook acquisition is usually called codebook training. Codebook training is an iterative process, and its core idea is the alternate use of the nearest neighbor principle and the best centroid principle. The nearest ...

AI Technical Summary

Problems solved by technology

The effect of directly using the codebook obtained by the traditional method in the quantization of spatial orientation information is not ideal, mainly because the codebook cannot reflect the subjective perception characteristics of the human ear well, and the subjective perception distortion of the quantization result is relatively large, which in turn affects the quality of 3D audio coding. Decoded overall subjective hearing

Images