Cramer-Rao Lower Bound for DOA Estimation with an Array of Directional Microphones in Reverberant Environments

Weiguang Chen,Cheng Xue,Xionghu Zhong
DOI: https://doi.org/10.21437/Interspeech.2021-2267
2021-01-01
Abstract:Existing direction-of-arrival (DOA) estimation methods usually assume that signals are received by an array of omnidirectional microphones. The performance can be seriously degraded due to heavy reverberation and noise. In this paper, DOA estimation using an array with directional microphones is considered. As the signal response varies over different DOAs, the magnitude information as well as the phase information can be employed to estimate the DOA. We first introduce the spherically isotropic noise field using directional microphones. The Cramer-Rao Lower Bound (CRLB) for DOA estimation is then derived and compared with that using omnidirectional microphones under different signal-to-reverberation ratio (SRR) environments. In addition, we extend existing steered response power (SRP), minimum variance distortion-less response (MVDR) and multiple signal classification (MUSIC) estimators for the DOA estimation using directional microphone arrays. Both CRLB Analysis and DOA estimation show that better DOA estimation performance can be achieved by using a directional microphone array.
What problem does this paper attempt to address?