Audio classification and implementation method based on reclassification

Abstract

The invention relates to an audio classification method for classifying audio signals before audio coding; the audio classification method includes primary classification and reclassification, and is characterized in that the reclassification includes the primary classification result is smoothened. Besides, the invention also discloses an audio classification device arranged on the front end of an audio coder for classifying the audio signals; the audio classification device comprises a primary classifier and a reclassifier, and is characterized in that the reclassifier comprises a smoothing module for smoothing the primary classification result. The method and device of the invention are used to correctly distinguish music and voice from audio signals. As the reclassification includes the smoothening on the primary classification result, the occasional false judgment caused by overquick audio type switching is eliminated, and the operational complexity is also educed so that the correct and simple distinguishing of music and voice is realized.

Description

technical field [0001] The present invention relates to a device for distinguishing whether an audio signal is speech or music before audio coding and its implementation method, more specifically, an audio classification device based on reclassification and its implementation method. Background technique [0002] [001] Voice and music are two main types of audio data, and the classification of voice and music is one of the important means to extract audio structure and content semantics. In addition to limited registration information such as sampling rate, quantization accuracy, and encoding method, original audio data itself is only a non-semantic symbol representation and unstructured binary code stream, lacking content semantic description and structured organization. How to extract the structured information and content semantics in the audio, so that the unordered audio data becomes orderly, is the key to the practicality of content-based audio retrieval technology. C...

AI Technical Summary

Problems solved by technology

The above classification methods have the following disadvantages: (1) they can only describe static statistical characteristics such as the mean and variance of the audio, while audio signal features usually have temporal statistical characteristics, for example, there are generally rhythms or drums that can reveal the theme in music; and In speech, unvoiced and voiced sounds tend to alternate, and these features are time-dependent

(2) Decision rules and search order are not necessarily optimal

The disadvantage of the above classification method is that the classifier needs to be trained with a large amount of data in advance, the whole process has a large amount of calculation, and it is not easy to implement in hardware

[0006] In addition, the prior art does not actually disclose the audio classification based on reclassification and its implementation method

Images