Research on Speech Quality Evaluation Method Based on Deep Neural Network

Abstract

With the continuous development of artificial intelligence research and high-performance computers, deep learning and related theories, neural networks have become mainstream acoustic models. This article starts with the overall framework and basic principles of speech recognition, and studies the language recognition system under deep neural network. Through deep neural network, it can better deal with the large amount of Chinese data processing process, reduce model complexity, and improve the accuracy of Chinese speech recognition rate. For this reason, this paper uses the deep neural network-based Chinese speech recognition system as the research object to study the speech quality after recognition. The results verify that the deep neural network language model has a better recognition accuracy.