Imitating Human Auditory Processing for Urban Soundscape Measurement

Intelligent measurement nodes offer the opportunity to perform advanced soundscape analysis, instead of just logging the sound pressure level. A model which mimics human auditory system is proposed and applied to analyze urban soundscapes. It is constructed as a combination of two types of neural networks: a Self Organizing Map that allows –after extensive training– to identify cooccurring sound features and a Locally Excitatory Globally Inhibitory Oscillator Network that is able to segregate them. The model takes into account the context of the listener and can be tuned to classify typical sounds of the soundscape at the location of the microphone.