Research On Perception Technology For Robot Based On Visual And Auditory Information Interaction

With the development of science technology, robot plays an important role in daily life, thus research on such visual and auditory perception robot is of great significance. Since human plays a critical role in environment information, detecting and recognizing human’s voice and visual image from environment has been of great significance. In this paper, human’s voice and facial image are taken as robot’s perception targets, and robot perception technology based on visual and auditory information interaction is studied, especially the sound source localization technology and face recognition technology. The mainly research work of this paper includes the following aspects:(1)A robot’s visual and auditory interaction pattern is studied which is inspired by human and takes human as a target. A visual and auditory interaction system is designed, includes sound detection module, sound source localization module, navigation module, target detection and recognition module, and wireless communication module. Each module interacts with each othet, so that to make the robot get abilities to perceive information.(2)Robot’s auditory perception technology towards environment is studied. Firstly, sound source localization based time difference of arrival algorithm is studied. Then the accuracy of time delay estimation method is improved through using voice activity detection, improving generalized cross correlation weighted function and doing interpolation. Finally, the Spherical interpolation location method is studied to make the localization more quickly. The experiment result shows that the improved algorithm can effectively improve the accuracy and robustness of sound source location.(3)Robot’s visual perception technology towards environment is studied. Face detection technology is studied and an improved method is put forward. Firstly, the face target area is detected. Then, this region is divided into different blocks. Next, wavelet decomposition and symmetry processing method is adopted to cluster similar data. Furthermore, the singular value decomposition method is used to make the feature extraction more efficiently. Finally, sparse representation is used to do face recognition. The experiment result shows that the improved algorithm has higher recognition rate.