Narrowband direction-of-arrival estimation for binaural hearing aids using relative transfer functions

State-of-the-art hearing aids feature multiple microphones per hearing aid and employ spatial filtering algorithms that are in general superior to single-channel algorithms. Spatial filtering algorithms require usually the direction-of-arrival (DOA) of the desired sound source. Recently developed informed spatial filters are based on a parametric signal model, and require an estimate of the DOA per time and frequency. In this paper, we develop a novel DOA estimation algorithm for hearing aids that estimates the narrowband DOA by comparing estimated relative transfer functions (RTFs) with anechoic RTFs per direction from a database. Furthermore, we propose a weighting based on the coherent-to-diffuse ratio to increase the robustness against estimation errors. The algorithm is compared to two existing narrowband DOA estimation techniques. One of the existing techniques is extended to use an arbitrary number of microphones to be able to detect DOAs in the full plane around the listener. Simulations show that the accuracy of the proposed DOA estimator is higher than the existing estimators in diffuse noise fields. Finally, we demonstrate that the proposed DOA estimator makes it possible to apply an informed spatial filtering technique to binaural hearing aids.