Research And Realization Of Home Robot Speech Recognition And Source Localization Technology

With the improvement of robot intelligence,a large number of robots have begun to enter the family and serve humanity.The robot auditory system can not only directly interact with human beings,but also can judge the location of sound sources in a special environment that cannot be seen or entered,and make up for the limitation of the limited visual range.It is an important direction for the intelligent development of robots.Based on the research of acoustic source localization technology based on microphone array,this paper proposes an inertial module-based robot sound source localization system model based on inertial navigation technology.The model first uses the microphone array to collect the sound source information,and uses the generalized cross-correlation(GCC)time delay estimation algorithm to obtain the time delay value to obtain the azimuth information.Then the robot moves slowly.The inertial module measures the acceleration and angular velocity information during the movement of the robot.After the quaternion is obtained by the angular velocity,the acceleration is transformed into a coordinate system.Acquire sound source information to obtain azimuth information.Finally,the location of the sound source is estimated by combining the azimuth information calculated by the robot after acquiring the sound source information twice and the displacement of the robot.After the simulation of the sound source positioning system based on the inertia module,a speech recognition and sound source localization system was designed and implemented on the hexapod robot platform.This text uses MEGASUN-M6 speech module to design the robot speech recognition system;Sound source localization system collects the sound signal from the microphone array,and after amplification and A/D conversion,sends the digital signal to STM32 through the wireless serial port module for time delay estimation.The displacement information collected by the inertia module JY901B finally shows the estimated sound source coordinate position on the OLED display System test results show that the recognition rate of the home robot speech recognition system can reach 81.67%within 5 meters;the average azimuth error estimation of the sound source positioning system is 12.4 degrees,show the average estimation error of the sound source distance is 7.6%.The positioning accuracy is in line with the system.Design goals.