基于深度学习的语音识别系统实现方法  

Implementation Method of Speech Recognition System Based on Deep Learning

作　　者：窦亚珍[1] DOU Yazhen(Henan Vocational College of Agriculture,Zhengzhou 451450,China)

机构地区：[1]河南农业职业学院,河南郑州451450

出　　处：《电声技术》2024年第10期74-76,共3页Audio Engineering

摘　　要：研究基于深度学习的语音识别系统实现方法,首先探讨语音识别系统的总体框架,其次深入研究梅尔倒谱系数(MelFrequency Cepstral Coefficient,MFCC)的提取和深度卷积神经网络(Deep Convolutional Neural Network,DCNN)的基本原理,最后基于Python和PyTorch框架进行系统测试。实验结果表明,所提方法在准确率、精确率及召回率方面均表现优异,能够较好地捕捉大多数样本。This article studies the implementation method of speech recognition system based on deep learning.Firstly,explore the overal framework of the speech recognition system.Secondly,delve into the extraction of Mel-Frequency Cepstral Coefficient(MFCC)and the basic principles of Deep Convolutional Neural Network(DCNN).Finally,conduct system testing based on Python and PyTorch frameworks.The experimental results show that the proposed method performs wellin accuracy,precision,and recall,and can capture most samples well.

关 键 词：深度卷积神经网络(DCNN) 语音识别 PYTHON 

分 类 号：X784[环境科学与工程—环境工程]