基于一维卷积神经网络的语音识别系统构建方法

扫码查看

原文链接

NETL
NSTL
万方数据

中文摘要：提出一种基于一维卷积神经网络(1D Convolutional Neural Network,1D-CNN)的语音识别系统.首先研究基于1D-CNN的语音识别系统框架,其次重点介绍使用TensorFlow构建该系统的方法,最后采用LibriSpeech数据集,在无噪声、轻微噪声和严重噪声条件下进行系统测试,并使用准确率、召回率、F1 等指标进行评估.实验结果表明,所提出的系统在无噪声和轻微噪声条件下具有较高的识别准确率和稳定性,即使在严重噪声环境中也表现出较好的健壮性.

外文标题：Construction Methods of Speech Recognition Systems Based on 1D Convolutional Neural Network

外文摘要：The article proposes a speech recognition system based on 1D Convolutional Neural Network(1D-CNN).Firstly,the framework of a speech recognition system based on 1D-CNN is studied.Secondly,the method of constructing the system using TensorFlow is emphasized.Finally,the LibriSpeech dataset is used to test the system under conditions of no noise,slight noise,and severe noise.Accuracy,recall,F1,and other indicators are used for evaluation.The experimental results show that the proposed system has high recognition accuracy and stability under both noiseless and slightly noisy conditions,and exhibits good robustness even in severely noisy environments.

外文关键词：

1D Convolutional Neural Network(1D-CNN)speech recognitionsystem constructionTensorFlow framework

作者：

刘洋、廉咪咪

展开 >

作者单位：

郑州工商学院,河南郑州 450000

关键词：

一维卷积神经网络(1D-CNN) 语音识别系统构建 TensorFlow框架

出版年：

2024

DOI：