Abstract
For the purpose of developing refined devices for environmental recognition in robotics, a practical method for 3D-localization of a stationary random acoustical source in near-field is proposed by introducing multiple fixed point-detectors. Though 3 points detection is sufficient for 3D-localization, it is essentially lacking in range resolution in the direction perpendicular to a plane determined by 3 independent detecting points. By setting one more detector out of the plane, and using three pairs of estimated auto- and cross-spectra of signals detected simultaneously at 4 fixed point-detectors, information on three sets of differences of propagation distances and the amplitude attenuation of wave propagation are derived in such a manner that the derivation is independent of shape of power spectra of detected random signals so as to be applicable to any stationary random acoustical source, and additive corrupting noises at detectors are canceled out as completely as possible. Then, with an aid of geometric relations of signal detection, the source position in 3D-space is estimated in the basis of the extracted information. After the principle of the proposed 3D-localization and the required conditions being made clear, the effectiveness of the proposed method as well as that by 3 points detection reported previously are experimentally evaluated under practical conditions in an acoustically treated room. The results of the experiments illustrate the effectiveness as well as fundamental characteristics of the proposed methods for 3D-localization.
