中国科学院自动化研究所机构知识库

Robot control based on audition plays an important role in many robot applications, which is generally implemented by three steps: aural features extraction, sound source localization and approaching control strategy. This paper presents a novel and practical auditory perception and control method adoptable to the small-sized microphone array. Firstly, a novel aural feature extraction strategy, namely dynamic signals matching (DSM), is proposed to search for the matching points of the signals with similarity measurement based on gradient and amplitude by dynamic programming. Secondly, a new sound source localization method called steered grid search (SGS) is proposed, which combines the source localization method based on time delay of arrival (TDOA) and spatial search strategy to achieve the robustness and real-time performance. Thirdly, inspired by image-based visual control, the interaction matrix is deduced and applied to transform the spatial information of the sound source into auditory control, based on which a mechanism for robot to approach the sound source is implemented. Finally, we apply our methods to a mobile robot equipped with a small array of six microphones. The experimental results show the feasibility and the suitability of our method under adverse conditions.