Research On Target Recognition And Localization Methods Of Mobile Robot Based On Visual And Auditory Fusion

With the development of people's life and information technology,the robot plays a more and more important role in family service.As human play an important part in living environment,this paper targets people,analyzing people's visual and auditory features,mainly studying the target recognition and orientation technology which is based on mobile robot integrating vision and audition.The main work of this paper is as follows:Firstly,this paper summarizes current situation of mobile robot's target recognition and orientation technology.The system platfonn for visual-auditory fusion of mobile robot is designed and the interactive control of mobile robot and the application of the scene are analyzed.Secondly,this paper researches target location of mobile robot.In the study of acoustic source localization,it brings GEVD to suppress noise impact in MUSIC.Mobile robot gains the azimuth basing on sound source localization and combines movement distance to figure out the sound distance.The experiment validates the accuracy and robustness of this method.Thirdly,this paper researches the technology of speaker recognition and face recognition.Ahead of face recognition,face detection is performed to locate the face region,then the face image is divided into blocks and the feature is extracted by wavelet decomposition and singular value decomposition.At last,sparse representation is used in face recognition.In speaker recognition,Mel frequency cepstrum coefficients of speech are extracted by homomorphic processing and cepstral analysis.Through the LBG clustering processing,we establishes the codebook model for the speaker.By the LBG clustering processing,we establishes the codebook model for the speaker.Experiments show that the validity of speaker recognition and face recognition.Fourthly,this paper researches the recognition technology based on the fusion of auditory and visual information.In the matching and decision layer,speech and face information are fused and recognition.The methods of speech and face weighting for speech priority matching and face priority matching are proposed.Compared with the non weighed fusion,the weighted fusion performs better in the experiment.In the decision level,the output of speaker recognition and face recognition is nonlinear weighted by the fuzzy integral and the experimental results show that the fuzzy integral is effective for audio-visual fusion.Lastly,this paper sets up the platform for mobile robot's target location and recognition on the basis of target identification and orientation.