The Research And Realization Of Phonetic Instruction Release Technology Based On DSP For Autonomous Carrier

With the rapid development of the society and the surging of all kinds of new technology, At present, higher demand and great expectations are stressed on all the products and devices which are typically featured with interactivity, convenience, comfort and intelligence. At the present time, the great challenges a large number of technicians are facing up with are how to break through the existing man-machine communication configuration so as for the audio recognition technology to be introduced into the routine management and the issue of the action instructions for the robotic equipment and home devices etc. This article and its relative researches are illustrated based on the above- mentioned matters.On the basis of a thorough analysis for its research focuses and its main problems in different stages of audio recognition and in combination with the advantages and disadvantages of all kinds of modern technology together with the specific characteristic of application for this project, An idea has been put forward in this article, that is, an experimentally environmental platform will be built by way of selecting single chip minicomputer on the basis of DSP chips used in the existing RSC-300/364 audio recognition system. No matter what kind of condition it is used, the audio recognition has to be faced with two cases, which are called audio signal extract and audio content understanding respectively.During the period of the audio signal extract, the differences, such as, environmental background noise, the quality of the microphone, the age, sex, personality, habit of dialect, spiritual status and the language types etc of the speaker is probable to influence and bind to the convenience and final effect of the build-up of the experimental system. In order to eliminate all these unfavorable factors, the author has not only illustrated many successful experiments in many other cases but also discussed the constraints of different vowels, consonants and frequency band etc on the adaptability of these methods in the adoption of noise removal by signal, noise removal by parameters and anti-noise modeling etc, furthermore, the corresponding strategies to solve these problems have also been mentioned. That is, the evaluation of the audio characteristic extraction, audio sequence programming mechanism and the audio spectral peak value; the determination of the audio linear predication parameters based on the minimum mean-square deviation as well as the effect on the accurate evaluation of audio parameters (such as reference volume frequency, resonance peak, power spectrum etc.), which is caused by audio signal waveform by using variable parameters in the case of low information rate. In consideration of improving the strength of the system, the accuracy of the conclusion, the reliability of the system and the constant improvement, the article has also showed relatively technical details accumulated from the previous successful recognition results etc. for the purpose of achieving nondeterministic speaker gradually.Audio recognition can not exist without the support of the physical hardware. According to the general objective and fundamental guiding principle of the project and on the basis of detailed illustration of the features of DSP chips used for audio recognition system RSC-300/364, many problems and its solutions which are commonly encountered in the period of designing experimental platform for single chip computer have been discussed thoroughly. Furthermore, the functional module and its final effect obtained from the adoption of assembly languages has also been mentioned in this essay. In the latter part of the essay, the operation and test conditions of the experimental system has been mainly introduced and the analysis on the existing problems and the orientation for future improvement have also been done concisely.