Research On Applying The Speech Recognition In Marine Simulator

As an important tool for navigation education and training,marine simulator has been widely used in recent years.However,the traditional interaction methods such as mouse and keyboard are used in the existing marine simulators.Sometimes,the user's attention is not completely focused on the training task,and part of the man-machine interaction experience of the operation function is not good enough.Speech recognition technology has been gradually used in intelligent products due to its advantages of natural use and convenience.But at present,the research on the application of speech recognition technology in the field of ships and navigation is relatively few.It is not suitable for the development requirement of the next generation marine simulator and intelligent ship.Aiming at this problem,this paper studies the key technologies of speech signal preprocessing,feature extraction,recognition decoding and speech recognition model.Based on the actual situation of navigation language and marine simulator operation,taking the maritime English as the material,a navigation text corpus was established through annotation and preprocessing,and the language model suitable for navigation was established for the first time by using the modeling method of N-gram statistical language model.The relevant corpus of speech is collected with the operating terms of marine simulator,and the audio corpus suitable for marine simulator is established by annotating the audio content and pronunciation.Using GMM-HMM and DNN-HMM modeling methods,the acoustic model is constructed by using phonemes as the modeling unit,and the corresponding acoustic model is trained by using the audio corpus of marine simulator as the training data.The speech model,the acoustic model and the pronunciation dictionary are combined organically to form a decoding network based on the weighted finite state machine structure,then the GMM-HMM speech recognition system and the DNN-HMM speech recognition system were built for the marine simulator.At the same time,in order to explore the application of large vocabulary continuous speech recognition in navigation,a large vocabulary continuous Chinese speech recognition system was constructed based on Chinese open source corpus in preparation for the further study of marine speech recognition system.On this basis,the speech recognition system built for navigation simulator is combined with speech synthesis to realize the speech interactive function of PC marine simulator(three-dimensional ship simulation platform),which enhances the user's interactive experience.