Design And Implementation Of Universal Voice AI Cloud Platform For Smart Home Speech Recognition

When it comes to the voice technology of smart homes,it should be emphasized that semantic recognition technology.In recent years,with the development of artificial intelligence,especially deep neural networks,the bottleneck of speech recognition in smart homes has been broken.In order to solve the research on the relative passive and lack of semantic analysis of the existing home control system in the selection of speech recognition platform,this paper combines artificial intelligence,deep learning,Deeplearning4j framework,Baidu speech recognition API,and stuttering participle to design a smart home speech recognition universal voice AI cloud platform,which not only allows users to choose the appropriate speech recognition engine,but also uses the neural network model to perform semantic analysis on the recognized text.According to the smart home device control instruction set,extract effective user data and control the smart home.Optimized to improve smart home control and enhance the user experience.The establishment of the intelligent home device control instruction set is the key to voice control.According to the research on common voice control commands for different people in smart home,this paper designs an instruction set that meets the needs of smart home control.On the basis of the completion of the instruction set,combined with the requirements of the cloud platform,the function design of the cloud platform mainly includes four functions:intelligent cloud middleware,semantic analysis,voice receiving and command sending.The middleware realizes the speech recognition function by calling Baidu speech recognition API;since the semantic analysis greatly affects the result of speech recognition,this paper chooses the word segmentation tool to process the word segmentation,and uses LSTM neural network algorithm to realize semantic analysis;The voice receiving module establishes a socket connection between the platform and the gateway,and realizes conversion of voice data from a JSON string to a voice object;The module encapsulates the instruction information,and establishes a socket connection to complete the instruction transmission according to the TCP/IP protocol.This paper adopts the platform-as-a-service software architecture in cloud computing,and combines the cloud platform overall framework diagram to develop the cloud platform using the"Spring+Spring MVC+MyBatis" technology framework,including user management,device management,voice reception,middleware,and semantics.Identify and command the sending module and implement the functions of each module.At the end of the thesis,the semantic analysis module of the platform is tested and the recall rate is 79.03%.The test results show that the semantic analysis model built in this paper can improve the accuracy of speech recognition and achieve the goal of perfecting smart home voice control.