The Design And Implementation Of Speech Interface Device Based On MOST Network

The car becomes more and more ordinary in people's life, and it becomes "Multi-Media Car" as the vehicle devices increase greatly. How to control these multimedia devices and the communication between them become more and more important. MOST is an effective way to solve the vehicle devices'communication problem. It is a new multimedia communication network which is flexible. And it is used by many different manufacturers as it can transmit audio signals synchronously. Because the traditional control system which is based on eyes and hands, can not keep the driver's hands on the wheel and eyes on the road, so it is not safe. And the speech control system can avoid these problems, so it is the direction of the vehicle human-machine interface.The paper is mainly about the design and implementation of the speech interface based on MOST network. The software design and hardware design of the speech interface are researched. The software design is based on the speech recognition theory. To meet the practical demands, RSC4128 is selected to be the main controller of the system and the peripheral circuit is designed. The communication between speech interface and MOST network is also researched and designed. All of this is the basic work for next research. The experiments verify the function of the speech interface, and the once recognition precision can reach 95%, twice recognition precision can reach 99%. So it can be applied practically.The first chapter describes the research status of the speech recognition, MOST and vehicle speech techniques, also the goal of the speech interface system. The speech recognition methods for dependent speaker with small vocabularies are researched early, and now there are already many perfect methods and products. MOST is a new vehicle multimedia bus which provides an effective control on audio, video and control information. The MOST Cooperation, which applies oneself to spread MOST, now contains more than 70 key component suppliers. But the MOST techniques research in our country just begins. So it is important for us to research speech interface based on MOST. The second chapter talks about the theory of the speech recognition. Based on the basic theory of speech recognition, the paper describes the pretreatment and the methods of the speech recognition. The pretreatment includes noise elimination, preemphasis, enframing, adding windows and endpoint detection. After that, the feature extraction is described, mainly about MFCC. Then the speech recognition method DTW, which is the main algorithm of the speech interface, is described. At last, the chapter talks about other speech recognition methods:Hidden Markov Model and Artificial Neural Network.The third chapter is the hardware design and implementation of the speech interface. First, it introduces the composition of the MOST network, and the main controller chip is selected according to the hardware requirements. Then the main character of the system controller RSC4128 is described and its clock and reset circuits are designed. Based on RSC4128, the entire system is designed and the peripheral circuits, which include power, keyboard, LED, program memory, data memory, MOST transceiver, microphones and speaker, are designed, too. The speech interface communicates with outer controller ATmega128 by serial port, and ATmega128 communicates with MOST controllers OS81050 by I2C bus. Then, the speech command is sent to the control channel of the MOST network by fiber. At last, the command is done by the actuator in the MOST network.The forth chapter is the software design of the speech interface. The program flow chart of the entire system is described. It mainly talks about speech recognition flow and speech training flow, particularly about how the speech interface sends control command after recognition. And then it describes other functions of the system, which includes how to delete the speech templates and how to set speech recognition level. At last, the critical data structures, such as the recognition result and the error code, are described. The error code, which storages the error information when the system runs, is used to interact with user by the prompt speech.The fifth chapter is the experiment part. First the chapter describes the picture of the system hardware, and then it introduces how to make speech command, human machine interaction sound and the communication protocol between speech interface and MOST network. After that, it describes the module tests, the system function tests and the integration tests with MOST network. The module tests verify that the LED module and the serial module can work well. According to the recognition precision experiment, which was made in quiet environment at normal speak rate, the system has a good recognition precision for dependent speaker. And the control ability test of the speech interface on the MOST network shows that the speech interface system can control the MOST network well. At last, the entire system function is showed, mainly about the training and recognizing process.The sixth chapter summarizes the work in the paper and the prospect we will do. The speech recognition for independent speaker and the integration of CAN-MOST gateway can be the further development of the speech interface. Based on the research of speech recognition theory for the dependent speaker and the speech recognition chip RSC4128, the speech interface device based on MOST network is designed and implemented successfully, and the experiment result shows that the speech interface has a good recognition for dependent speaker, and a good control ability on MOST system. It meets the practical demands.