A Model-Based 2-D Head-Tracking Method Using Microphones at the Ears

Head-tracking is essential to dynamically reproduce binaural signals. If a crosstalk cancellation system is used for reproduction via loudspeakers, the head-tracking can be conducted based on acoustic delay estimates. Instead of estimating the delays between anchor sources and microphones at the listener's ears, we propose to obtain these delay estimates by continuously measuring the impulse responses between the loudspeakers and the microphones. Based on the measured delays, a nonlinear least-squares problem is presented to jointly estimate position and orientation in the horizontal plane. This problem formulation incorporates a novel angle-dependent parametric delay model to increase tracking precision. We show that our presented parametric delay model is superior to the free field and Woodworth's delay model in terms of head-tracking precision. Assuming free field conditions despite the presence of the listener's head is demonstrated to causes a systematic error. Moreover, we validate the proposed head-tracking method in a preliminary dynamic evaluation in an adaptive crosstalk cancellation system.