Research And Realization On The Recognition Technology Of Speech Based On DSP

The Language is the most natural and the most convenient form that the mankind correspond by letter each other, so the speech communication is a kind of ideal correspondence method between operator and robot. To make robot have the person's intelligence really, it is essential to have the speech correspondence between operator and robot. Therefore, the research work that carries on the speech control has the actual meaning and the application foregrounds.The thesis mainly pays attention to the design of software and hardware circuits, it introduced the analytical and processed foundation knowledge of the speech signal first. And then, the fundamental theory of speech recognition is discussed.The paper also introduces how to design the system, in which TMS320VC5402 is used as its core circuit. In the hardware circuit TI Company's DSP is used as controller of signal collection and processor of digital signal. The TLC320AD50C accomplishes the transformation of A/D and D/A of the audio signal. The communication module transfers information between the audio control system and robot by RS232-bus. The logic function module accomplishes all the logic function of the system hardware by CPLD chip.Allow for the feature of mandarin speech, we apply the existing arithmetic to design the software system, and the design process is described in this part. The process of speech recognition goes along as following: pre-emphasis, frame blocking, windowing, draw the characteristic of speech signal, pattern matching. Linear prediction coefficients analysis is used to draw the characteristic of speech signal. Because the number of the system's command is less than 50, we choose the dynamic time wrapping to be the algorithm for speech recognition.