The Design And Study Of Speech Recognition Based On Humanoid Robot Of Gd-1

With the development of science and anthropology, people crazily want to find an industrious robot as computer working for people even it can enter into people’s feelings. Recently with the fast development of humanoid robot, people enjoy the highly humanoid appearance more and more. While the humanoid robot can’t have a massive progress because of it can’t listen and speak as humankind. The thesis which comes for the reason makes it possible that the humanoid robot can listen and speak due to PC and makes the corresponding movement.The thesis realizes the speech recognition system of Speak-Independent, it adapts the template matching method based on Dynamic Time Warping (DTW) algorithm. In order to make the system strong and robust, the thesis chooses eight proper vocabularies, As for Speaker-Independent, a method is proposed to train the speech templates, which is based on the fusion of overall mean and variance information. Maximum Likelihood Estimation (MLE) method is mainly applied in the thesis in order to make sure the rightness of the recognition.In the whole system, the thesis uses the Multimedia API of Microsoft to collect the speech signal of real-time online, realizes the speech recognition based on MATLAB, realizes the serial communication based on RS485. The robot can make a corresponding movement due to the speech command of PC. In the environment of lab, the rate of Tightness can reach beyond90%. While the point which the thesis should improve is that the robot system can’t get rid of PC. which the thesis should continue to work for.