Language identification method and system

Abstract

The method and system for speech identification is disclosed, which picking up vector sequence of characteristic parameter from speech signal; obtaining at least one candidate word after speech identification; segmenting speech signal into speech segmentations corresponding to verification unit; determining vector sequence of characteristic parameter of speech segmentation for verification; carrying out verification to generate score of verification. In the method, score of verification is calculated by using quasi-neural network for verification units. The quasi-neural network is perceptive from multiple layers. Score of verification is obtained by feedforward operation under comparison between vector sequence of characteristic parameter for verification and multiple layers of perception. Adding scores of verification of all speech segmentations obtains score of verification for word. Determining whether the candidate word is accepted or rejected based on defined threshold level.

Description

technical field [0001] The present invention relates to a method and system for verifying words, especially a method and system for verifying words with high reliability, which can be applied in a noisy environment. Background technique [0002] Press, word verification (utterance verification) technology is used to verify the correctness of a candidate word obtained by a speech signal through a speech recognition bee, so that the system can act according to the verified candidate word. For example, in a voice dialing system, when the system receives a voice signal containing a telephone number, it will recognize and verify the numbers contained in the voice signal, so as to perform dialing according to the verified number. [0003] Among the known word verification techniques, the most widely used methods mainly include two categories, respectively using decoding technology for word verification and word verification methods based on hypothesis testing. [0004] see figur...

AI Technical Summary

Problems solved by technology

However, the above-mentioned verification method based on hypothesis verification must establish a verification model and a reverse model for each recognition unit; and the verification of each speech segment must go through two verifications, which will greatly increase the system load

[0006] In addition, the above-mentioned interpretation methods and hypothesis testing methods are mostly implemented in an environment without noise interference. Therefore, once they are applied in a noisy environment, it will seriously affect the accuracy of word verification and reduce the reliability of the recognition results, and then lose the speech recognition system. affinity

Images