Design Of Speech Recognition System Based On Android Platform

In recent years,with the continuous development of mobile terminal service and artificial intelligence matures,and with the increasingly stringent requirements on the subjective feelings on the product,natural language interaction product demand is growing.Android is an important operating system for mobile terminals,based on this platform to achieve speech recognition is of great significance.Most of the current speech recognition technology is based on simulation,failed to achieve the application of the project.The current mobile terminal speech recognition function is based on third-party platform to realize,such as Google,Baidu and so on.There are two ways: online speech recognition and offline speech recognition,the drawback of the online speech recognition is that it must be connected to the network.The drawback of the offline speech recognition is that the users can not customize the content to be recognized,both of which are depend on the processing of the speech signal by the third party.If build up a separate speech recognition system,and the content can be developed by the developer for the user,and in the case of offline to realize speech recognition,mobile terminals can reduce the use of network resources,and it will greatly improve the user experience.In this thesis,a lot of research on the related technology of speech signal processing has been done,including preprocessing,feature extraction,and so on.The following solutions are proposed on the basis of these theories: First,analyze HMM and RNN,the basic theory and the application in speech recognition.Then,the advantages and disadvantages of HMM and RNN are analyzed,and the design scheme of HMM and RNN is proposed.Finally,the Android platform speech recognition system is designed,the influence of adjusting training parameters on the system is analyzed,and compare to other speech recognition systems by a large number of experimental tests,evaluate the advantages of the recognition rate of speech recognition system and anti-noise performance Finally,the existing problems of the system and prospect the future.