Abstract:
Acoustic Vector Sensor (AVS) arrays have been studied extensively for underwater applications, but very little work has been done for applications in air and especially for speech. AVS’s are extremely useful for picking up of audio signals from both azimuth and elevation planes. AVS’s are compact compared to other microphones arrays with same number of sensors. The most important aspect of the AVS compared to that of other microphone arrays is, they can measure scalar acoustic pressure and all three components of acoustic particle velocity [1]. Applications of an AVS include localisation and multichannel enhancement of speech sources.
The directional information form the AVS’s can be extracted with ease, using the ratio of the average energies of the signals, how ever this is not as straight forward as it appears. In this paper we will try to propose a design change to AVS’s that will minimize shadowing by others sensors while the maintaining the compactness of the AVS and maintain a polar pattern that is constant for wide band of frequencies. This will include a description of the polar response of two different AVS constructions as well as results showing the performance of the AVS for localisation of speech sources. Results will be compared to those obtained from a Uniform Linear Array (ULA).
Authors: M. Shujau, Dr. C. Ritz, Dr. I, Burnett
Event: SF08: Search and Information Extraction from Audio Data Workshop