| Speech recognition technology develops rapidly in recent years. It starts from laboratory to market and flows into the smart home, intelligent wearable devices, intelligent vehicle and other fields. At present, a lot of attention is focused on driving safety. Speech recognition technology is applied to vehicle system so as to achieve that the vehicle speech can control navigation and air conditioning, which increases the convenience of driving vehicle operation system, avoids the driver’s attention dispersed and ensure driving safety.Based on the project of Chinese speech real-time recognition method in noisy environment which is from the provincial science and technology department’s development project, this paper studies the driving language Chinese speech recognition method after removing the noise of the quiet environment--that is the small-vocabulary recognition algorithm for specific people method. This paper analyzes the characteristics of voice commands’ spectrogram and uses compressed sensing to write programs,which achieves a new speech recognition method based on the spectrogram characters. This approach effectively cuts down the amount of data processing and quickens the processing speed. At the same time, it also ensures the accuracy and timeliness of the on-board system identification instructions.This thesis mainly puts to use MATLAB7.0 software to research the recognition algorithm, programming and simulation. Firstly, the author preprocesses the speech signal and unifies the quantization standard to generate spectrogram with the same size. Secondly, it aims to use the sparse algorithm to turn a spectrogram into a sparse matrix and find the right voice commands template. Once again, this paper makes the identification of the corresponding sparse matrix testing instruction in the sparse domain with the template.Eventually, this study calculates the recognition rate of each instruction and verifies whether this method is reasonable. |
PDF Full Text Request