Applications
A sound recognition system for the real world
One goal of our research is the development of a microphone that learns to describe its acoustic environment.
This system should work always and anywhere and should be able to describe sonic environments in a similar
informative way as humans can. This system is not unlike a bind person who awakes in an unknown place and listens
to learn the properties of the environment. Such a person might describe the environment as “A quiet place. Somewhere
indoors. Inside the house in hear the sound of 2 or 3 people walking and talking at a distance. They do things, and
are maybe cleaning the house. Outside I hear with the sound of the wind rustling in trees, some birds, and an occasional
car passing at low speed. All in all this is a quiet and relaxing place.” Systems like this can be used to describe
sonic environments of all sorts. For example the report above can be used to describe the the acoustic quality of a house
that is for sale, or it can be used to compare the sonic ambiance before and after infrastructural measures (new sound barrier,
traffic circulation measures, etc.).
Modeling auditory attention and auditory annoyance
We do not hear all sounds. Or more precisely, we do become only aware of a minority of all audible sounds. That is
important because most sounds are not at all important and generally we do want to be bothered by these sounds. However,
sometimes sounds have properties that make it almost impossible to ignore them. For example very loud sounds or sounds that
are otherwise significant such as your name of the the gentle buzzing of a mosquito. When irrelevant sounds attract your attention
persistently, they become irritating, annoying, and eventually even a health risk. We want to model these processes for societal use.
Physics-based sound analysis
Sound events are produced by physical processes. A passing car produces the sound of passing car and not the sound of a violin
because its physical properties do not allow it to produce anything else. The same holds for a violin. The question then is “What in the
sound of a violin identifies it as a violin?” We assume that sounds do originate that the different sound producing processes
structure the sound in predictable patterns. Before the sound reaches an ear transmission effects might perturb it and mix it with
other sounds. It is the tasks of our auditory system to discover the patterns imposed by the source and interpret the