Single sound direction estimation for mobile platforms

Abstract

This is a study of a real-time single sound source direction estimation system for moving platforms. A modified explicit adaptive algorithm for estimating the time delay between the signals captured at the microphone sensor array using wavelet transform is proposed. An active microphone sensor positioning method is also proposed for increasing the estimation performance and reducing computational loads. An explicitly adaptive time delay estimation (EATDE) method is suitable for fast tracking of a time-varying delay, but the conventional EATDEs may fail to converge to a real delay parameter if the distance between sensors are too wide or the acoustic signal bandwidth is too broad. A modified EATDE using Haar wavelet transform is proposed in order to avoid this convergence failure. The theoretical analysis and numerical simulations show that the proposed algorithm converges globally to an unbiased delay estimate. A fast convergence speed and small estimation variance can be achieved using wavelet scale adaptation. This method can also be applied to remove the bias error when there exists secondary sound sources. Active microphone positioning method is to change the relative position between the sound source and microphone sensors so that a calculation of geometric position using the time delay information should not be necessary. The precision of position estimation can also be improved because the sensitivity of time delay with respect to the relative position is maximized and the dependence on speed of sound is diminished. A microphone placement in 3-dimensional space are also devised. Finally, the experiment system with a steerable microphone array which consists of 4 microphones and 2 stepping motors are developed and the result of estimation performance tests in real environment are provided.