Research On Smart Home Design Based On Speech Recognition

With the development of information technology and the increasing degree of importance of the Internet of Things of our country,people's lives have become more comfortable and intelligent.In this process,smart homes have sprung up.Smart home take apartments as medium,which manages home equipment centrally.It is convenient,comfortable,safe and environmental friendly.It can bring intelligent and fashionable life to the family and improve the living quality of residents.At present,most smart home devices use mobile phones as the main control terminals.The control mode of "Wi-Fi" + 4G,which is still a traditional control method,is used to complete the control of smart home devices,cannot satisfy users' the variety needs of control for smart homes.In recent years,as the booming of artificial intelligence and the unceasing maturity of machine learning technology,voice recognition has been rapidly developed,and the recognition accuracy has been improved endlessly.In order to meet the higher needs—more convenient and efficient—for home equipment control,the combination of smart home and voice recognition technology has been an inevitable trend.Therefore,this paper proposes a deep learning speech recognition intelligent home control system based on AllJoyn framework and noise reduction automatic encoder.The speech recognition model is used to parse out the phrase control instructions to achieve the purpose of home equipment control management.The deep learning speech recognition model of this paper mainly consists of two parts.The first part is unsupervised learning pre-training.Randomly change some network node states before unsupervised pre-training,artificially simulate noise data,and then train each hidden layer sequentially by RBM weight matrix,and optimize by comparing the deviation of input data and output data.The purpose of the parameter.The second part is the supervised parameter adjustment,that is,the trained parameters are taken as the initial values of the whole network,and the back propagation algorithm is used to optimize and adjust the entire network model.The experimental results show that the speech recognition system is significantly improved compared with the speech recognition system based on convolutional neural network,both on the speech recognition rate and the robustness to noise.Combining the speech recognition model with the smart home system,the home control command is recognized from the phonetic phrase,and the human-computer interaction non-contact and convenient control has been realized,thereby making the system more in line with people's requirements.The smart home gateway and the mobile App independently integrate the voice recognition system.The smart home control solution uses the AllJoyn framework as the basis of LAN communication to realize the interaction between the intranet intelligent device control and the mobile App.The external network uses the intelligent voice gateway as the link and the voice gateway integrates AllJoyn.The framework establishes a long connection with the server through WebSocket.On the one hand,the voice signal is received through the AllJoyn framework to control the smart device,and on the other hand,the control command sent by the mobile terminal reaches the gateway via the server,and the LAN device interacts through the gateway.After the laboratory scene test,the feasibility of the smart home control system based on speech recognition was verified,which effectively improved the intelligence of the smart home.