Prof. Dr. Georg Martin Klump
Dr. Ulrike Langemann
Adapting to a noisy world: Processing of auditory signals
Almost any natural environment is noisy, thus noise commonly affects the perception of auditory signals and may seriously impede acoustic communication. As a consequence of turbulences in the atmosphere, amplitude fluctuations are often superimposed upon environmental sounds when they are transmitted. Furthermore, many natural sound sources produce temporally structured signals, e.g., a group of humans chatting or birds singing in a dawn chorus (see example of waveforms in Fig. 1).
Figure 1: (A) Waveform of a dawn chorus recorded in a middle European mixed oak forest. (B) Synthesizing a dawn chorus by digital mixing of the song of four bird species. (C) Imitating the natural situation in the laboratory experiment using coherently amplitude-modulated noise.
In behavioural experiments with trained starlings (Fig. 2) we investigate sensory mechanisms that can make use of temporal fluctuations in the envelope of background noise and other environmental sounds. These mechanisms may be used for the perceptual grouping of signal components, segregation of signals from background noise and hence in the analysis of auditory scenes. Acoustic components that were emitted from the same sound source or that have been transmitted together usually carry coherent modulation patterns (i.e., modulations that are correlated and in phase). Vertebrates have evolved mechanisms to exploit these modulation patterns. If maskers carry coherent amplitude modulations in different frequency bands, signal detection may be considerably improved by up to 20 dB (i.e., a factor of 10). This masking release has been termed "Comodulation Masking Release".
Figure 2: European starling (Sturnus vulgaris)
As part of the DFG-supported research association "Auditory Objects" we investigate Comodulation Masking Release not only in the avian auditory system (i.e., in the starling) but also in mammals like the Mongolian gerbil. Additional research in profoundly hearing impaired human subjects with cochlear implants intends to add to the knowledge necessary for optimizing signal processing of speech processors used for cochlear implants.