Speaker confirmation method and speaker confirmation device used in short voice condition

Abstract

The invention relates to a speaker confirmation method and a speaker confirmation device used in a short voice condition. The method comprises the steps of extracting a linear prediction cepstral coefficient, a Mel frequency cepstral coefficient and a Delta characteristic from an objective voice signal; combining the linear prediction cepstral coefficient, the Mel frequency cepstral coefficient and the Delta characteristic, thereby obtaining a plurality of effective characteristic vectors; reducing the dimension number of the effective characteristic vectors by means of a partial fuzzy PCA method; and modeling by means of the dimension-reduced effective characteristic vectors according to a Gaussian mixture model, and identifying the speaker of the objective voice signal. Compared with the prior art, the speaker confirmation method and the speaker confirmation device are advantageous in that a single characteristic is replaced by a characteristic combination, thereby improving number of dimensions of the effective characteristics and compensating defects of a characteristic sample, and furthermore the local fuzzy PCA is used for performing effective dimension reduction on the combined characteristic, thereby reducing time-space complexity of the system under a precondition of low effect on identification rate.

Description

technical field [0001] The invention belongs to the technical field of voice recognition, and in particular relates to a speaker confirmation method and device under short voice conditions. Background technique [0002] In the process of speaker recognition technology moving toward practical application, the following situations are often encountered: 1. Terrorists or intercepted objects, for the consideration of anti-interception, often speak very short, sometimes even only two or three words. Although text-related techniques are effective for short speech speaker recognition, text-related speaker recognition cannot be used for such situations. 2. Due to the problem of the call transmission channel, the quality of many phone calls is very poor, resulting in serious voice interruption. The usual solution is to remove discontinuous speech segments that hardly contain or can hardly extract speaker information, and the result must be that the effective speech becomes shorter. ...

AI Technical Summary

Problems solved by technology

Although text-dependent techniques are effective for short speech speaker recognition, text-dependent speaker recognition is not applicable to such situations

2. Due to the problem of the call transmission channel, the quality of many phone calls is very poor, resulting in serious voice interruption

The usual solution is to remove discontinuous speech segments that contain little or no speaker information, and the result must be that the effective speech becomes shorter

3. In the process of solving the multi-speaker problem, due to the lack of maturity of the current voice segmentation technology, or the existence of overlapping voices, it is necessary to It is cut off before being sent to the recognizer, which will inevitably lead to shorter effective speech

4. In some commercial occasions, users are unwilling to speak a long voice

[0013] At present, the GMM system is the most commonly used system for speaker recognition technology. This system is based on the theory of statistical models and requires training and test speech to reach a certain length. Otherwise, the system performance will be greatly reduced.

In other words, in the case of short speech, the system recognition rate will be greatly reduced

Images