Determinants of accuracy in a terrestrial microphone array

Abstract

Acoustic sensor arrays can allow researchers to localize the position of vocalizing animals. During the course of research on a threatened bird species, the greater sage-grouse, we developed a 24-channel wired array to non-invasively monitor male courtship displays at traditional display grounds (i.e. leks). Here we describe a study in which we localized repeated playbacks of four local species while varying speaker position, the number and arrangement of microphones, and accuracy of speed of sound and sensor location estimates. As expected, localization accuracy was lowest when the speaker was outside the array and when using a linear microphone arrangements. We found no overall effect of species identity in spite of strong differences in time and frequency structure of the playbacks, although we did find significant interactions of species with other factors in our analysis. Simulated errors in speed-of-sound-in-air and estimation of sensor position revealed that while localization was most accurate when these errors were small, localization was still possible even with relatively large errors in these two factors. While we hope these results will help researchers to design effective sensor arrays, specific outcomes will depend on study-specific factors as well as the specific sound processing and localization algorithms employed.