Interspeech'2005 - Eurospeech

Lisbon, Portugal
September 4-8, 2005

Self-Organizing Chirp-Sensitive Artificial Auditory Cortical Model

Luis Weruaga (1), Marián Képesi (2)

(1) Austrian Academy of Sciences, Austria; (2) Technical University of Graz, Austria

This paper presents a novel signal-processing-based artificial model of the auditory mechanism. This work is inspired by the psychocortical fact that the biological cochlea is very sensitive to frequency-varying tones, or chirps. The method uses a novel combination of several (at least three) Harmonic-Chirp transform instances, that project the time-frequency energy on different views; all projections are data-mined by self-organizing unsupervised layers of Radial Basis Functions, this process being driven by the Expectation Maximization algorithm. The mechanism shows biological parallel, such as intrinsic chirp sensitivity and response to the logarithm of the stimulus energy. The proposed model is validated with several mammal sounds, such as human speech, bat echo, and several aquatic mammals.

Full Paper

Bibliographic reference.  Weruaga, Luis / Képesi, Marián (2005): "Self-organizing chirp-sensitive artificial auditory cortical model", In INTERSPEECH-2005, 705-708.