Beamforming with collocated microphone arrays

Abstract

A collocated microphone array, including three gradient microphones with different orientations and one omnidirectional microphone, was used to acquire data in a sound-treated room and in an outdoor environment. This arrangement of gradient microphones represents an acoustic vector sensor used in air. Beamforming techniques traditionally associated with much larger uniformly spaced arrays of omnidirectional sensors are extended to this compact array (1 cm3) with encouraging results. A frequency–domain minimum-variance beamformer was developed to work with this array. After a calibration of the array, the recovery of sources from any direction is achieved with high fidelity, even in the presence of multiple interferers. SNR gains of 5–12 dB with up to four speech sources were obtained with both indoor and outdoor recordings. This algorithm has been developed for new MEMS-type microphones that further reduce the size of the sensor array.

Acoustic vector sensor: reviews and future perspectives

“…For performance enhancement, a spatially displaced sensor array is usually used instead of a single sensor [2]. Recently, a new device named acoustic vector sensor (AVS) which consists of four sensors mounted at the same point is introduced [3–14]. Comparing with conventional sensor array, an AVS shows advantages in small size and light weight.…”

“…Existing research of the AVS is mainly theoretical [3–13, 15–25]. The pioneering work on source localisation with a single AVS is presented in [3].…”

“…Although fruitful achievements have been reported, only a few of them focuses on real recorded signals analyses [8–10] due to lack of low cost hardware devices for AVS. The interference cancellation on real signal collected by an AVS constructed using one Knowles pressure sensor and three directional sensors is investigated in [8, 9]. The speech enhancement performance on an AVS using the minimum variance distortionless response (MVDR) beamformer and Wiener post‐filtering is studied in [10].…”