Speech recognition method based on convolution neural network

Abstract

The invention provides a speech recognition method based on a convolution neural network, which is more good at extracting high-level features, has simple modeling process, is easy to train, has better generalization performance of the model, and can be more widely applied to various speech recognition scenes. The method comprises the following steps: S1, preprocessing the input original speech signal; S2, extracting the key feature parameters reflecting the characteristics of the speech signal to form a feature vector sequence;S 3, base on that DCNN network model, taking the connected time classifier CTC as a loss function, constructing an acoustic model of an end-to-end mode; S4, training the acoustic model to obtain the trained acoustic model; S5, inputting the feature vector sequence to be recognized obtained in the step S2 into the trained acoustic model to obtain a recognition result; and S6, performing a subsequent operation on the basis of the recognition result obtained in step S5, that is, obtaining a word string capable of outputting the speech signal with a maximum probability, that is, a language character after the original speech is recognized.

Description

technical field [0001] The invention relates to the technical field of speech recognition, in particular to a speech recognition method based on a convolutional neural network. Background technique [0002] In speech recognition technology, the GMM-HMM (Gaussian Mixed Model-Hiddden Markov Model) model has always occupied a dominant position as the acoustic model of speech, but due to the characteristics of the GMM-HMM model itself, the GMM-HMM acoustic model is First of all, the alignment operation needs to be performed, and the data of each frame needs to be aligned with the corresponding label. The alignment process is cumbersome and complicated, resulting in a long training time, and because the model is a combination model of GMM and HMM, the specific modeling process is relatively simple. It is complex and has certain limitations in the specific application of speech recognition technology. Contents of the invention [0003] In order to solve the problems of long tra...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problems of long training time, complex modeling process and limited application of existing acoustic models, the present invention provides a speech recognition method based on convolutional neural network, which is better at extracting high-level features, and the modeling process is simple and easy to train , The generalization performance of the model is better, and it can be more widely applied to various speech recognition scenarios

Images