基于DNN-LSTM模型的智能家居语音识别系统设计
Smart Home Language Recognition System Design Based on DNN-LSTM Model
林勇升 1田美艳 1王鑫2
作者信息
- 1. 厦门华天涉外职业技术学院,福建 厦门 361000
- 2. 福建师范大学,福建 福州 350007
- 折叠
摘要
为提高智能家居语言识别系统的准确率和匹配率,通过在深度神经网络(DNN)模型的第1 层增加长短时记忆神经网络(LSTM)结构,运用信息熵实现对声学训练与语种匹配,设计了基于DNN-LSTM模型的语音识别系统.将该系统应用于语音识别,结果表明系统的中英文声学模型识别准确率为 96.6%,语种匹配准确率为 95.8%.该系统对提升智能家居的智能化水平具有一定的实用价值.
Abstract
In order to improve the accuracy and matching rate of the smart home language recognition system,a speech recognition system based on the DNN-LSTM model is designed by adding a short-term memory neural network(LSTM)structure to the first layer of the deep neural network(DNN)model and using infor-mation entropy to achieve acoustic training and language matching.The system is applied to speech recogni-tion,and the results show that the accuracy of Chinese and English acoustic model recognition is 96.6%,and the language matching accuracy is 95.8%.This system has certain practical value to improve the intelligent level of smart home.
关键词
语音识别/DNN-LSTM模型/智能家居Key words
speech recognition/DNN-LSTM model/smart home引用本文复制引用
基金项目
福建省教育厅教育提质培优项目(ZS22051)
出版年
2024
NETL
NSTL
万方数据