面向智能家居设备控制的唇形识别系统设计
Design of the Lip Recognition System for the Smart Home Equipment Control
宋隆霖 1张培卿1
作者信息
- 1. 博西华家用电器有限公司,安徽 滁州 239000
- 折叠
摘要
本文指出传统的智能家居系统多通过语音指令的方式对设备进行控制,难以满足嘈杂环境、听力障碍人员的应用需求.为此,本文将唇读技术应用于智能家居设备控制,设计了唇形识别系统,通过对识别用户的唇形,实现对设备控制指令的识别.该系统以单目摄像机为视频输入,设计唇部定位、运动特征提取、特征时序建模等模块,通过对用户唇形的识别,实现了部分智能家居控制指令的识别.实验结果表明,在本文场景下,通过将三维卷积与时间卷积神经网络相结合,系统识别精度可达 78%.
Abstract
This paper points out that as for the smart home system,the equipment is always controlled by voice,in case of noisy environment,and hearing-impaired users,such control can hardly be implemented effectively.In order to solve the problem,this paper applies the lip reading into the smart home,designs a lip reading system.Through the recognition of the lips,the control introductions can be achieved.More specifically,the system adopts the camera as the input,designs the lip localization,motion feature extraction,and feature time modeling.With these modules,control introductions of the equipment can be recognized by extracting the lips of users.It can be improved that the recognition accuracy can achieved to 80%by incorporating the 3D convolution with the time convolutional neural network.
关键词
智能家居/唇形识别/三维卷积/卷积神经网络Key words
smart home/lip recognition/3D convolution/convolutional neural network引用本文复制引用
出版年
2024
NETL
NSTL
万方数据