A Planer Moving Microphone Array for Sound Source Localization

Sound source localization (SSL) provides service robots with the ability to hear like humanity, which makes sense in complex dark indoor environments. Since vision can not always workable while audio can do. From the data collection view, the larger the number of microphones mounted on a robot, the better result of SSL can reach. But one large number of microphones array, say 16 microphones array, would occupy a huge volume of the robot. Thus, we designed a framework using 4 planar moving microphones structure to mimic a 16-microphone array for sound source localization to save space. Due to special design, the proposed 4 planar moving microphones structure can form different patterns to do 3D SSL, including angle, distance and height. Besides, we also build circular 6-microphones array and planar 4x4-microhones array for experiment comparison, both of which can rotate around for fair. The experiment is conducted in a normal classroom. The result shows that the proposed framework could reach generally 80%-90% accuracy for angular estimation and about 85% accuracy for distance, height estimation of planar 4x4-microhones array.