Design Of Vehicle Maintenance Assistant System Based On Speech Recognition

With the development of social economy,the number of cars in China is increasing rapidly.Under this background,doing a good job of vehicle detection and repair is of great significance to ensure the safety of the car.In the traditional operation process of vehicle maintenance enterprises,the maintenance results are recorded by manually filling in paper forms,which causes low maintenance efficiency,high maintenance cost,and difficulties in saving maintenance results.To solve these problems,speech recognition technology can be introduced,it converts the speech spoken by maintenance workers into text maintenance results.After that,an information system for results management can be applied,and the paperless vehicle maintenance process can be realized.Therefore,the design of a vehicle maintenance assistant system based on speech recognition is conducive to efficient and intelligent vehicle maintenance,which has engineering application value.This thesis researches the speech noise reduction technology for the noisy vehicle maintenance scene firstly.On the basis of DCU-NET noise reduction model,the attention mechanism is added,the upsampling process is optimized,and an improved DCU-NET is proposed.The PESQ score of speech processed by improved model reaches 3.11 at 15 dB SNR.Secondly,this thesis researches the speech recognition technology for vehicle maintenance scene.By analyzing the speech recognition task,a speech data set is made.On the basis of the BiLSTM-CTC model,whose recognition accuracy is better than that of traditional GMM-HMM model,GRU is used to replace LSTM and the BiGRU-CTC model is constructed,which optimizes the real-time performance of speech recognition.The recognition accuracy of BiGRU-CTC model is 95.7%.Combined with speech noise reduction and speech recognition technology,this thesis finally designs and implements a vehicle maintenance assistant system.The system adopts the hybrid architecture of C/S and B/S,including several modules such as maintenance management subsystem,maintenance client subsystem and server.It provides services for maintenance administrators and maintenance workers,which can realize the speech recognition for vehicle maintenance scene and has the function of auxiliary maintenance.The test results show that the function and performance of vehicle maintenance assistant system meet the expected requirements.It can be applied in practical scene to help vehicle maintenance enterprises improve operational efficiency.