From active, non-linear processes in the ear and numerical hearing models to music transmission on the Internet with MP3 or objective assessment of the speech quality of mobile phones: Hearing acoustics have a direct impact on our daily lives - not only when a hearing impairment occurs or when you can no longer understand anything at a lively party. Analysing the effective function of hearing as a complex overall system is interesting for physics. Translating this analysis into a hearing model enables a wide range of technical applications.

Auditory scene analysis: How do we form an acoustic picture of our surroundings?

Hearing works unconsciously; hardly anyone realises the complex processes in the ear and brain that turn sound waves into "heard" information. One of the greatest mysteries is how we filter out the voice of the person who is saying something from a multitude of sound sources - people talking, dogs barking, noisy cars. This works perfectly for healthy people, but it's a completely different story for people with hearing loss: They can only communicate if the background noise is largely eliminated. Hearing aids only help partially because it has not yet been possible to technically reproduce the complicated processes in the auditory system for object formation. A cocktail party: voices, clinking glasses, subtle music. Everyone is talking to each other, sometimes in pairs, sometimes in small groups - but not everyone understands what the other person is saying. Some people frantically try to read what is being said from the movements of their lips - but to no avail, their hearing and brain are overwhelmed in this acoustic environment. 15 per cent of all Germans suffer from sensorineural hearing loss; the trend is rising, if only because our society is getting older and hearing loss is a classic problem of old age. In young and healthy people, the auditory system works perfectly and, above all, in continuous operation.

Digital hearing aids

The first hearing aid with digital processing of acoustic signals was introduced in 1996. Although digital technology had already found its way into every living room with the CD at this time, it was a surprise to everyone in the hearing aid industry that it was possible to accommodate this technology and this processing power in the smallest possible space in the hearing aid and to operate it with the lowest possible energy consumption. Not even mobile phones, which are generally regarded as miracles of innovation, achieved this level of efficiency. The introduction of the digital hearing aid can therefore easily be described as a quantum leap. Since then, development has progressed rapidly and all major hearing aid manufacturers now offer fully digital devices that allow complex digital processing of acoustic signals. The available processor performance in hearing aids is increasing at about the same rate as that of home PCs, which is mainly due to the packing densities of the components used on the processor modules, which are increasing equally in both areas.
Computers (e.g. hearing aids) are not yet able to replicate the capabilities of human hearing. A hearing aid that amplifies all acoustic signals equally is of no use in cocktail party situations. Instead, it must separate the sound objects and amplify them selectively. In addition to the high selectivity of the auditory system for different frequencies/pitches, the perception of amplitude modulations (rapid volume fluctuations) and directional hearing play a key role in this object separation and recognition. Directional hearing, in turn, is closely linked to binaural hearing, which enables the sound waves arriving at both ears to be offset.

What we are working on ...

In Oldenburg, we are working on algorithms that bring the cocktail party effect to hearing aids. We are working on several approaches to noise cancellation - from (binaural-connected) directional microphones to stochastic algorithms.
Our work involves basic research in the field of algorithms for speech processing in hearing aids. Among other things, we are working on algorithms

Hohmann, Volker
Ewert, Stephan Dietz, Mathias Grimm, Giso
Fredelake, Stefan Klein-Hennig, Martin Eilers, Reef Harlander, Niklas