Binaural Direction-of-Arrival Estimation Incorporating Head Movement Information

Binaural direction-of-arrival (DoA) estimation algorithms typically assume that the recorded signal is captured by a stationary head. If the head moves, its orientation changes and DoA algorithms may require time to readjust to the new coordinate system. However, modern devices such as headphones and hearing aids, incorporate sensors that provide information about the head movement. In this contribution we propose the new movement-invariant beamforming (MIBF) binaural DoA estimator. It is based on the target beamforming (TBF) DoA estimator and incorporates head movement information to improve DoA estimation performance in dynamic situations. We evaluate TBF and MIBF algorithms on binaural recordings including head movement data. Our results show that the MIBF estimates the source direction more robustly during head movements. Additionally, it is demonstrated that the inclusion of motion data via MIBF aids in resolving the issue of front-back confusion.