Impulsive sound source localization using peak and RMS estimation of the time-domain beamformer output

Abstract

This paper presents a beamforming technique for locating impulsive sound source. The conventional frequency-domain beamformer is advantageous for localizing noise sources for a certain frequency band of concern, but the existence of many frequency components in the wide-band spectrum of impulsive noise makes the beamforming image less clear. In contrast to a frequency-domain beamformer, it has been reported that a time-domain beamformer can be better suited for transient signals. Although both frequency- and time-domain beamformers produce the same result for the beamforming power, which is defined as the RMS value of its output, we can use alternative directional estimators such as the peak value to enhance the performance of a time-domain beamformer. In this study, the performance of two different directional estimators, the peak and RMS output values, are investigated and compared with the incoherent measurement noise embedded in multiple microphone signals. The impulsive noise source is modeled as a triangular pulse, and the beamwidth and side lobe level of the time-domain beamformer are formulated as functions of the pulse duration, the microphone spacing, and the number of microphones. The proposed formula is verified via experiments in an anechoic chamber using a uniformly spaced linear array, and the results show that the peak estimation of beamformer output determines the location with better spatial resolution and a lower side lobe level than RMS estimation.