Based Dsp55x Series Of Speech Recognition System Design And Research

With the development of information society, the communication of man and machine is increasingly valued by the society. Speech recognition technology become one of the important research-field of modern computer technology, it involves the linguistics, the computer science, the signal processing, and has the extremely broad application prospect.In the world there has a large number of people with disabilities, their life has a lot of inconveniences. This paper considers the difficulties of using arm-impaired people use such as: computers, TV remote control, air conditioning equipment commonly used remote control so when the family is not convenient. We imagine such a device: the speaker on the apparatus in the oral order, an order by the speaker apparatus of the treatment, to the speaker's purpose. Instruments to achieve this objective are designed in this paper based on DSP 55X series of speech recognition system platform. With this system through the equipment, to meet the needs speech processing and recognition occasions to use.The paper introduced the speaker recognition system based on DSP TMS320VC5509A, expatiated the composition of system, presented the working process of the system. The main process of speech recognition includes pretreatment, the vertex examination, the characteri- stic parameter extraction and the pattern matching. Pretreated mainly uses the pre-emphasis, the vertex examination uses method of VUS arithmetic, the main characteristic parameter of speech recognition uses Mel frequency cepstrum coefficient (MFCC), the recognition uses artificial neural network (ANN). Software system was used for the compilation of language and language C.The core circuit of system is TMS320VC5509A, with some peripheral for example TLV320AIC23. The sampling and A/D transformation of original speech are completed by TLV320AIC23, the speech signal is processed by TMS320VC5509A chip, internal storage is use of storage program date, the external storage is use of storage speech signal date. The system , on the result of test, real time, effectiveness and advantageous value, which can work in general environment. First, system was passed the test of TMS320VC5509A; second, the arithmetic was used in development of hardware and got the result of examplement.Through the simulation of the speech recognition algorithm used in this system, the effectiveness of the algorithm has been proved and the testing on hardware platform also gets the expected effect . Experiment finds that the speech recognition system designed in this article has a better recognition rate for isolated and unfilled words. Through extension for hardware systerm this system can be desiged for multi application speech recognition system.