Direction of arrival estimates using matching pursuit
Yasuhiro Oikawa, Yoshio Yamasaki · 2005
Many topics on microphone array processing, e.g. signal enhancement, noise reduction and direction of arrival (DOA) estimation, has recently been researched. Many methods has been suggested for DOA estimation and the number of estimated DOAs is generally limited by the number of microphones and we has spatial aliasing problem based on relation between the microphones distance and wave length. We has proposed new DOA estimation method using matching pursuit algorithm, which can estimate more DOAs than the microphones number. The DOA estimation we did consists of the following steps. First, we calculate the normalized power of the array output, P (θ), for each frequency bin using the Delay-and-Sum method. We then average P (θ) over all frequency bins. Finally, we perform peak-picking using a matching pursuit algorithm to estimate the DOA over all frequency bands. The matching pursuit algorithm includes, after each iteration step, a reoptimization of all DOAs found thus far. Its main characteristics is that it is possible to find true DOAs when the number of sources exceeds that of microphones. In this paper we report DOA estimation in real-world using this method. We could estimate more DOAs than the number of microphones. Our experiments yielded a better DOAs estimation with the new method than with the conventional method.