The Cocktail Party Problem

Even though modern sophisticated hearing aids can detect speech in noisy environments and be individually customized, the problem of filtering out unwanted information and noise in large gatherings remains a significant challenge.

This problem reflects the biggest challenge that the Hearing Systems group researchers work with, namely to explain the human ability to focus and understand speech from an individual talker in complex, noisy environments, commonly referred to as the Cocktail Party problem. To evaluate such situations, humans use not only the acoustic information from both ears but also visual information. The healthy hearing system is remarkably good at focusing on a single target sound while ignoring background noise. This ability is often substantially reduced in hearing-impaired listeners and cannot be fully regained even with state-of-the-art hearing instruments. However, there are also many people with normal audiograms that have problems listening and communicating with other people in noisy environments, such as at a restaurant or a cafe, especially if the acoustics are annoying.
In order to solve this problem, we need to create realistic conditions like a listener would encounter in a cocktail party or at a restaurant. Since field studies are time consuming and hard to reproduce, we are working to bring the acoustic scene into the lab. To achieve this an audiovisual laboratory is being created with many speakers and a giant screen that will surround the listener. In this way, we can create an accurate simulation of the real world, which we can use to evaluate and test modern hearing aids in a systematic and controlled manner.

The virtual reality environment will consist of a mixed-order ambisonics sound field reproduction system and a large curved screen or 3D-glasses. Once operational, the virtual reality environment will be evaluated by characterizing the perception of audio and audio-visual stimuli. Finally, audio-visual integration mechanisms of the human auditory system will be investigated with the focus on cocktail-party scenarios. This research is expected to lead to a better understanding of the audio-visual integration when listening in challenging situations. In this way, we will come a step closer to solving the cocktail party problem.