The Research Of Voice Processing And Recognition Algorithm Based On DSP

Today, with the rapid development of computer technology, society has entered the high-speed information age, Traditional information storage and dissemination of people's daily lives have been unable to meet demand. To make life faster and more convenient, people eager to speak can be resolved through a number of matters, such as voice dialing, voice access systems, voice input and so on. Meanwhile, with the digital signal processor and voice recognition algorithm development, and makes speech recognition a reality, Speech recognition will be applied to a broader range. Therefore, the speech recognition system has a very high research and practical value.In this paper, a DSP-based human-specific small-vocabulary isolated word speech recognition system, recognition system described in detail the hardware design process, Analysis and improvement of speech recognition algorithms, and simulation results are given voice recognition.Firstly, a brief speech signal basic theory of knowledge and relevant acoustic principles, four aspects of speech recognition of the basic issues from the speech signal preprocessing, feature extraction, training and match these are discussed.Then, focus on research and design of DSP-based Speech Recognition System Hardware Design. TMS320VC5402 as a hardware core of the entire system, peripheral circuits, including voice signal acquisition module TLV320AIC23, power circuit, LCD display, memory expansion, JTAG interface circuit.. The driver related system hardware is given. System discussed speech recognition system based on DSP software design, system threshold comparison method with endpoint detection, focus on the 24-dimensional Mel frequency parameters (MFCC) and dynamic time warping (DTW) identification of efficient algorithms.Finally, on the Matlab2006 of PC,DTW speech recognition algorithm simulation design has been given,then describes and analizes simulation environment and simulation results.