Auditory Model-based Tracking of Mixed Acoustic Sources

M. Képesi (Austria)


Speech processing, Auditory modeling, Pitch tracking.


Pitch tracking of acoustic sources in single-channel recordings is a difficult task, particularly when more than one dominant source is present, and when the sources have similar spectral characteristics and their pitch values are very close. It is interesting to note that the human auditory system is capable of tracking multiple acoustic sources, and also able to concentrate on any one of them. This paper tries to answer the question how can we segregate pitch trajectories of two or more speakers speaking at the same pitch. We introduce an improved auditory model with changing sensitivity based on the harmonic structure of the source we are listening to.

Important Links:

Go Back