Robust speaker distinguishing method based on multifactor frequency displacement invariant feature

Abstract

The invention discloses a frequency displacement invariant feature extracting method considering multifactor in voice, and the method is used for distinguishing text-independent speakers under a complex environment. The method comprises the step of: in consideration of the time, frequency, scale and phase information of voice, performing multifactor characterization on a voice signal energy spectrum through two-dimensional plurality wavelet transform of different scales and phases, in consideration of the displacement invariant feature of the frequency, calculating a displacement invariant feature projection matrix on a frequency order through a convolution-type non-negative tensor decomposition method to obtain a multifactor sparse feature, decorrelating the feature through discrete cosine transform, and calculating the first order and second order difference coefficients of the feature, thus finally obtaining speaker feature with good robustness. The frequency displacement invariant feature extracting method considering multifactor in voice disclosed by the invention calculates the multifactor frequency displacement invariant feature with robustness through the convolution-type non-negative tensor decomposition method to distinguish the text-independent speakers under a noise environment, so that good distinguishing accuracy is good.

Description

technical field [0001] The invention relates to a feature extraction method for improving speaker identity discrimination performance, and belongs to the technical field of speech signal processing. Background technique [0002] With the continuous development of computer and artificial intelligence technology, various intelligent machines participate in human production and social activities, so how to improve the relationship between people and these machines and make it easier for people to manipulate machines It is becoming more and more important, and language is the best way for people to communicate with machines. [0003] Speech signal processing is an interdisciplinary subject combining linguistics and digital signal processing technology. It is one of the important means of computer intelligent interface and human-computer interaction. Speaker identification is an efficient means of human-computer interaction. Its characteristics are signal The collection is conve...

AI Technical Summary

Problems solved by technology

[0017] At present, the characteristic coefficient of the signal in the text-independent speaker identification system can get better recognition accuracy in a relatively quiet environment, but when the surrounding environment is complex and there are many interferences and noises, the speaker identification performance will be reduced. The main reason is that the robustness of single-factor features (such as spectrum) is poor, and the training model of the speaker model does not match the test data.

Images