Student Theses - Details
Investigations on Loudspeaker-based Adaptive Binaural Reproduction
Themengebiet: signal processing, system identification, modeling, machine learning
Kategorie: Bachelorarbeit (BA), Masterarbeit (MA)
Status: Offene Arbeit
Tools: Matlab, Python, C/C++
The buzzword spatial audio is on everyone's lips these days - whether it's for surround sound playback in movies, playback of 360° recordings on YouTube, or spatial playback in FaceTime conferences. Most often, binaural signals are played back via headphones, but playback via loudspeaker-based systems also offers advantages...
During playback via loudspeakers, in contrast to headphone playback, sound also travels from the right loudspeaker to the left ear and vice versa. Therefore, crosstalk cancellation (CTC) is required in this case. For this, the transfer functions between loudspeakers and ears are needed. These can be estimated continuously in loudspeaker-based adaptive binaural reproduction with microphones placed close to the ear canal entrance to allow dynamic movement of the listener, while the crosstalk cancellation remains effective continuously. In addition, the position/orientation of the listener can be estimated from the set of estimated impulse responses (head tracking) to adjust the playback signal.
In this field, multi-faceted thesis topics can be offered in consultation, including these areas:
- system identification: adaptive filters, supervised machine learning, system modeling
- filter design
- investigations on system architectures
- head-tracking algorithms
- realtime implementations to build a demonstrator
- acoustic measurements for objective evaluation of a prototype system
An exact topic can be agreed upon in consultation and taking into account the student's own interests. The scope and difficulty can be adapted so that both bachelor's and master's theses are possible. If you are interested, don't hesitate to get in touch in order to discuss further details in a personal meeting.