Design And Implementation Of A Voice Search Control System On Smart TV

Benefit from the current stage of rapid development and technological innovation in the field of artificial intelligence,technologies such as speech recognition and natural language processing are widely used in various electronic products.Smart TV as one of the important products,compared to traditional TV can only through the remote control buttons to control,because of carrying the intelligent voice search feature makes it greatly increases the user experience of the product.Smart TV let users use microphone for voice input,and then go through speech recognition and semantic analysis and other technologies,finally returned to the user search results you need.In order to realize this function,it is necessary to build a voice search platform for practical application,so as to meet the user's voice search needs.The overall implementation of the system adopts the front-end JSP page +Java background + algorithm model,mainly using Java and Python language development,from the function of the software requirement analysis,system design,system implementation of a detailed description the voice interaction system.The main work includes: detailed analysis on the development of smart TV and voice interaction technology development of perfect and functional requirements of the users of TV function,and how the function of smart TV's voice interaction system can meet the needs of voice interaction.The system adopts B/S architecture,includes front-end page calls,Java back-end server calls the algorithm model for automatic speech recognition(ASR)and named entity recognition(NER),provide web client interface and display themselves.Carry on detailed system realization plan design: The front page is responsible for implementing the voice input,identification text and correct the text display;Java back-end implementation provide front page call API,accept the front page input voice and text,Save every time the voice file and corresponding text label to the database,and provide the page view function,for the convenience of later update and maintenance;The model of ASR and NER was trained by using recurrent neural network,ASR realizes therecognition of the input voice to get the corresponding Chinese text,and NER realizes the semantic analysis of the Chinese text to accurately locate the search content.The system is developed in strict accordance with the required specifications of each step.In practical application,the system runs stably and can achieve the use effect under the actual situation,speech recognition and semantic analysis,and other functions,can satisfy the user when using smart TV voice search control requirements.