基于语音识别实现智能家居的控制

北京理工大学珠海学院2020届本科生毕业论文Control of Smart Home Based on Speech RecognitionAbstractArtificial intelligence is one of the areas of most concern for researchers.With the adventof 5G era,the development of the Internet of things has stepped onto the highway.And withpeople's life is the closest contact,contact time is the longest household goods.All kinds ofnew technology into the field of smart home,make smart home more and more understandlife.Under the background that the pace of modern people's life is speeding up and theconsumption upgrade pays attention to the sense of experience,compared with the traditionalhome equipment that requires buttons and other interactive ways,the advantage of intelligentinteraction is becoming more and more obvious,which makes people love it more.At present,artificial intelligence and Internet of things technology is still in the stage of development andprogress and maturity,among which voice recognition and control of home functions is animportant development direction.The application of dynamic time warping(DTW)algorithm and hidden markov model(HMM)technology makes the theoretical basis of current speech recognition system moreand more perfect.Has nonlinear artificial neural network,the characteristics of the learningand adaptive,in speech recognition technology has the very good application effect.As neuralnetwork recognition algorithm for a single word recognition system research,the study ofneural network model is improved at the same time,improve the identification accuracy ofthe system,can realize many people voice recognition.According to the domestic and foreign research status in the speech,we simplified thehousehold control part,in speech recognition technology as the main body part,completed thehousehold control solutions.The system consists of four parts (audio preprocessing module,speech recognition model,audio acquisition module,network communication module).Research design in the process of the two plans to build artificial neural network model toperform speech recognition task,the small batch audio files to train the model using themicrophone recording equipment,using the trained model to simulate the audio recorded bythe recognition and the result will be sent via the socket network communication from onedevice to another,to achieve the result of the control signal,realized the speech recognitioncontrol simulation of smart home applications.2