Locating Multiple Incoherent Sound Sources in 3D Space in Real Time 2011-01-1667
A model based approach is developed to track and trace multiple incoherent sound sources in 3D space in real time. This technology is capable of handling continuous, random, transient, impulsive, narrowband and broadband sounds over a wide frequency range (20 to 20,000 Hz). The premise of this technology is that the sound field is generated by point sources located in a free field. To locate these sound sources, iterative triangulations are used based on the signals measured by a microphone array. These signals are preprocessed through de-noising techniques to enhance signal to noise ratios (SNR). Unlike the conventional beamforming, the present technology enables one to pinpoint the exact locations of multiple incoherent sound sources simultaneously by using the Cartesian coordinates, including sources behind measurement microphones. In other words, the microphone array need not face a test object, which is required in the beamforming. Only four to six microphones are used in this approach, with four being the bare minimum to produce the (x, y, z) coordinates of a target, and six being capable of producing more accurate and stable results. The microphone positions are reconfigurable to suit the test environment, provided that they are not on the same plane and their coordinates can be determined after reconfiguration. Experimental validations are demonstrated. Tests are conducted in various non-ideal environments such as inside a machine shop, hall ways of a large building, and crowded rooms where background noise and reverberation effects are relatively high. Both stationary and moving sources that produce arbitrarily time-dependent acoustic signals are used to test the effectiveness of source localization of this technology. Practical limitations of this technology are examined and discussed.