Bachelor Theses - Details

Motion Data-Supported Impulse Response Estimation

Betreuer:Tobias Kabzinski, Christoph Weyer

Themengebiet: signal processing, system identification, modeling

Kategorie: Masterarbeit (MA)

Status: Offene Arbeit

Tools: Matlab, ggf. C, C++

Task description

When reproducing a binaural signal with headphones, the binaural signal for the left ear and the binaural signal for the right ear exclusively reaches the left or right ear, respectively. When reproducing using loudspeakers signal from the right speaker also reaches the left ear and vice versa. In this case, a crosstalk cancellation (CTC) is required in order to only feed the desired signal to the respective ear. In an adaptive crosstalk cancellation system, the four transfer functions (cf. Figure), which depend on the listener and their position/orientation, are estimated using adaptive filters and microphones at the listener's ears.

In this thesis the core research question is to analyze how motion data can help to improve estimating the transfer functions and hence support the dynamic equalization. Therefore, models shall be developed which could possibly be integrated in the existing Kalman filter structure.

Firstly, the equalization of a single moving microphone shall be considered. Thus, the current transfer function to the microphone needs to be inverted. When the microphone moves, the transfer function changes and the equalization filter does not match the current transfer function any more. When the estimate of the transfer function is updated, the equalization filter must be updated, too. To predict the change of the transfer function (in the direct sound part) the microphone will be equipped with a motion sensor. The motion data shall be utilized to predict the change in the transfer function. A Kalman filter (or extend Kalman filter) could be used to track the motion and jointly estimate motion and transfer function. The developed methods shall be verified in simulations as well as in acoustic measurements (if time permits). Depending on the progress, the algorithms might be extended to multichannel (MIMO) systems.

If you are interested, feel free to contact me and we'll schedule a personal appointment.