Method for quickly recognizing speaker

Abstract

The invention provides a method for quickly recognizing a speaker and belongs to speaker recognition methods. The method comprises the following steps of: combining a Gaussian mixture model, and taking the supervector of the Gaussian mixture model as the feature parameter of the speaker; taking the supervector of the Gaussian mixture model as input, designing a one-class support vector machine classifier; and training N classifiers corresponding to N speakers, thus obtaining a voice sample of one speaker from one classifier. By utilizing the method, the speaker recognition speed is increased; for every new registered speaker, only one one-class support vector machine classifier is trained for the new speaker, so that the speaker recognition system has good extensibility.

Description

technical field [0001] The invention relates to a speaker recognition method. Background technique [0002] Speaker recognition, also known as voiceprint recognition, is a biometric technology that distinguishes speakers by voice for identification and authentication. At present, the speaker recognition method based on support vector has become a mainstream speaker recognition method. [0003] For the support vector machine, it is a two-class classifier. When it is applied to the speaker recognition system, it needs to complete the transformation from two classes to multi-class classification. The two-class support vector machine classifier can implement multi-class classification in two ways, one is a one-to-one implementation method. This method is a more commonly used multi-class classification method. The support vector machine multiclass classifier adopted in the patent CN1787075 and the patent CN102201237A is implemented in this way. For one-to-one multi-class clas...

AI Technical Summary

Problems solved by technology

Although the one-to-many implementation method is simple and intuitive, it needs to construct a small number of two-class support vector machines, but it uses all other class samples as negative samples, resulting in an imbalance of positive and negative samples, too large negative training samples, and longer training time

Especially when new categories are added, the negative samples change, and all support vector machines need to be retrained, resulting in poor scalability of the system

Images