Research Of Speech Recognition And Man-Machine Interactive System Of Domestic Robot

Language is a human faculty. As a survey shows, about 75% of the human communition is achieved by spoken language in everyday life. There is no doubt that language is one of the most acceptable ways to achieve man-machine interaction.In this paper, the method of extracting characteristic vectors is firstly introduced by an example. The implementation process of the HMM theory and DTW theory in speech recognition is also discussed after that, analyse is given concerning the difference of spectrogram between male and female. Then the function design of man-machine interactive system in domestic robot based on RSC4128 is introduced, For the hardware platform development, special funtions has been designed: powerdown sleep mode to conserve power; double input signals channels to support audio wakeup; two independent oscillators provide frequency; external 4Mbit Code Space and 2Mbit Data space for memory configuration and PWM to directly drive a loud speaker. For the Speech independent recognition sets,"T2SI"technology is used to create HMM-based SI recognition templates. For good speech recognition performance, board layout and microphone placement is carefully designed to bring highly recognition accuracy. For the software development section, programs flow chart which emphasiz the use of one or more speech technologies in an application is introduced.Finally, a series of experiments have been done to evaluate the performance of the man-machine interactive system which includes: Speech Independent recognition accuracy testing; different Speaking Style recognition accuracy testing; Speech Dependent recognition accuracy testing and System function demonstration. Analyse of the influence of different factors proves that the system can bring a highly recognition accuracy over 90% under normal noises environment and read style speaking. And also improvement of the system robustness has been proved during the recognition testing in different noisy situations and movement control experiment, which absolutely satisfys the requirement for normal domestic robots.