摘要
针对目前市场上语音控制家居产品的语音识别必须在服务端进行,在缺乏互联网情况下则无法使用,以及智能家居设备采用有线组网方式导致升级维护成本高的问题,设计一款采用WiFi、蓝牙等无线混合组网技术搭建的异构家庭网关,可以动态添加设备,组网灵活方便且成本低,即使在断网情况下,网关也可自动切换到本地语音识别。为此,使用基于上下文相关的GMM-HMM算法训练声学模型与语言模型搭建语音识别引擎,采用BP神经网络学习算法实现语音智能控制。测试结果表明,系统功能基本符合预期,在断网情况下系统仍可以进行语音识别并对家庭设备进行语音控制。
For the voice recognition of the voice control home products on the market,the voice recognition must be identified on the server side,and it cannot be used in the absence of the Internet.The smart home equipment adopts the wired networking method,which leads to the problem of high upgrade and maintenance cost.The heterogeneous home gateway built by the bluetooth wireless hybrid networking technology can dynamically add devices.The networking is flexible and convenient,and the cost is low.Even in the case of network disconnection,the gateway can automatically switch to local voice recognition.To this end,the context-based GMM-HMM algo- rithm is used to train the acoustic model and the language model to build the speech recognition engine,and the BP neural network learning algorithm is used to implement the speech intelligent control.The test results show that the system function is basically in line with expectations,and the system can still perform voice recognition and voice control of the home device in the case of network disconnection.
作者
张珣
李道磊
ZHANG Xun;LI Dao-lei(College of Electronics and Information,Hangzhou Dianzi University,Hangzhou 310018,China)
出处
《软件导刊》
2019年第5期49-52,57,共5页
Software Guide
关键词
智能家居
语音识别
异构网关
智能学习
smart home
speech recognition
heterogeneous gateway
intelligent learning
