Design And Implementation Of Speech Recognition System Based On The Train Vehicle Environment

Safety is the eternal theme of railway operations. With the perfection and improvement of the railway system, the requirements put forward for the construction of train traffic safety monitoring system becomes higher and higher. Regulated speech of train drivers is an important part of the train regulation, the implementation of intelligent management, identification and evaluation of drivers’voice operation is imperative for train safety systems research. In this paper, the voice work process for train drivers launched an in-depth study, design and implementation of speech recognition for train drivers and operating conditions of the train-car speech recognition evaluation system. In this paper, the research was divided into two parts, the first part was the algorithm designed for train drivers of speech recognition, the second part was summary and development of the first part. It’s the core part of the thesis, mainly to complete train-vehicle speech recognition system. This paper took train drivers belonging to a particular driving field as our object. Around the speech signal forming mechanism, speech features, feature extraction and template establishment of train drivers, this paper carried out a complete and through theoretical and experimental research. Based on these, an accurate, real-time and robust train vehicle speech recognition system was designed and realized. The specific contents are as follows:(1) Based on the general characteristics of drivers’speech, this paper analyzed the formation mechanism of drivers’speech and the basic processes of speech recognition. Besides, this paper made a research and improvement on pre-emphasis, windows adding and feature extraction of common method.(2) Based on the particularity of train drivers, sound environment characteristics of train-vehicle, effects of interaction on the voice signal, this paper analyzed the design and algorithms for train drivers’endpoint detection algorithm, noise minus algorithm, the template library establishing algorithms and the trend cancellation algorithm. In addition, this paper made an appropriate improvement on the basis of summing up experience of the research work today It provided the basis for the development and design of the train vehicle speech recognition system.(3) Against the train drivers’speech signal, according to the analysis of the specific operational working purposes of train inspectors, the speech recognition algorithm suitable for train vehicle was determined. Then, this paper made a deep analysis of the working features of existing train monitoring and recording device combined with the included part of TAX device. In this way, the control signal of the speech recognition module was gained to get the specific operating time and recognition content of the speech recognition module in this system. It improved the efficiency and reliability of the system.(4) The output module of the speech recognition result was designed. On the next step, a several of the means to output the results was provided. The issue achieved the specific hardware and software design based on microcontroller and FPGA. Finally, this thesis analyzed the shortcomings and deficiencies of the system after summarizing the entire issue in order to provide a direction for the further improvement of the system.