Construction and Measurement of an Endfire Microphone Array Brent Kirkwood Duration: 1 Month Full Time (January 2002) Supervisors: Mogens Ohlrich Finn Jacobsen In cooperation with Brüel & Kjær Sound & Vibration A/S Download Full Report (PDF) (400 KB)

Abstract This report describes the construction and measurement of a small endfire microphone array that can be used in applications where it is desired to isolate a voice or sound coming from a single direction. It has been shown by the Laboratory of Acoustical Imaging and Sound Control that microphone arrays are a very effective tool to improve speech intelligibility in noise (Merks, p. ix). Such arrays could find application in assistive hearing devices, hands-free telephones, speech recognition systems, intercom systems, or any system in which it is desired to detect speech in an environment with competing undesirable signals. By increasing the signal-to-noise ratio (SNR), intelligibility can be likewise increased, whether that intelligibility is measured from a human's perspective or by a computer's speech recognition accuracy. Automobiles, aircraft, cocktail party-like environments, or any noisy environment in which the space required for implementation is also an issue, are possible environments for application of such a device. The frequency range from 180 Hz to 4500 Hz has been shown to be important for speech intelligibility in noise, with 1 kHz to 3 kHz being the most important band (Merks p. 39). Typical telephone bandwidth is from 300 Hz to 3400 Hz. This report describes an array designed and constructed to focus on these frequencies. The design of the system described here dictates that it should be used with a minimum operating distance of 6.8 cm. For many applications this restriction is minor, however it would be important to keep this minimum distance in mind if one were to select an application. The array described in this report can be built with simple and inexpensive analog electronics. No digital signal processing should be required to obtain similar results to those presented here, however digital signal processing could be employed to further improve the results. The array processing is primarily based on the Ph.D. thesis work of Dr. Ivo Merks at Delft University of Technology in The Netherlands, however some variations have been made in order to expand on Merks' work. As an application test, it was desired to build this array using Brüel & Kjær silicon-based 1/4" MEMS Multi-field microphones, Type 4945. For simplicity and speed of project completion, processing was performed by computer instead of analog electronics.