International Graduate School for Neurosensory Science and Systems
University of Oldenburg
"Signal Detection Enhanced by Comodulated Noise"
Many natural sounds share the property of having common amplitude modulations across different frequency regions. Examples of these kinds of sounds are the masking noises used for experiments concerning comodulation detection differences (CDD) and comodulation masking release (CMR). Typical stimuli for these experiments consist of one or more masking noise bands and a signal that is either a pure tone or a noise band as well. Previous research in humans and other vertebrates has shown that in CDD and CMR experiments signal detection thresholds vary depending on the correlation structure of the envelopes of the constituting noise bands. Based on neurophysiological data using similar stimuli as in psychophysics, a simple model that is capable of largely reproducing psychophysical signal detection thresholds in response to CDD and CMR stimuli is proposed. The model is based on peripheral processing and has the following stages: The incoming stimulus is frequency filtered, its envelope is extracted, followed by a compressive nonlinearity. In a last stage the temporal average is computed. Signal detection is realized by registering changes in the mean compressed envelope of the filtered stimulus. It turns out that the reason for threshold differences between signal detection in comodulated noise and signal detection in incoherently modulated noise is the interference phenomenon in combination with the compressive nonlinearity.