A Speech Recognition Method Based on Neural Network Model

Abstract

The invention discloses a speech recognition method based on a neural network model, comprising: obtaining a training sample collection, extracting the first characteristic parameter of the training sample collection, obtaining a neural network model that can be used for recognition, and establishing a template library; obtaining prediction samples Collect and input the neural network model, extract and obtain the second feature parameter, match and predict the second feature parameter with the first feature parameter in the template library, and obtain the recognition result; wherein, the neural network includes sequentially set convolution Neural Networks and Recurrent Neural Networks. Compared with the prior art, the speech recognition method provided by the present invention effectively combines and connects the convolutional neural network and the cyclic neural network, which can increase the overall learning efficiency and robustness of the network while ensuring the correct rate of speech recognition. Improve speech recognition performance.

Description

Technical field [0001] The present invention relates to the field of speech recognition, and is specifically a speech recognition method based on a neural network model. Background technique [0002] People in today's world have reached a very high degree of application and dependence, on this basis, people have also been more eager to operate more intuitively to complete their own purposes, if they can directly adopt sound propagation The form of directly controlling the machine to help you do what you want to do, will make a lot of things half a matter, so speech recognition technology operations. Voice recognition technology The nature is to process and classify the information contained in the sound, and people can identify, and the machine obtains information contained in the sound. The birth of this technology makes some simplification of the machine's operational order, making people's hands to be liberated, greatly facilitating our lives and work. Neural networks are an i...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to overcome the poor performance of neural network speech recognition in the prior art, and provides a speech recognition method based on neural network models, which effectively combines convolutional neural networks and cyclic neural networks to ensure that While increasing the accuracy of speech recognition, it increases the overall learning efficiency and robustness of the network, and improves speech recognition performance

Images