Ongoing Doctoral Thesis
Auditory motivated signal sampling and representation techniques:
Application to efficient analog-to-information conversion
Sound is a stimuli which has been studied scientifically and modeled mathematically.
The efforts have been quite successful when we look at the way our voice is transmitted for
communication or music processing is done in studios based on these principles. However, the
many tasks such as speech recognition, language learning, sound mixture segregation remain open
problems for machine implementation and also speech processing through cochlear implants is far
from satisfactory. We attempt to learn from the way mammalian auditory system performs signal
processing and pattern recognition. The biological signal processing aspects are nonlinear and
do not in general fall into the paradigm of linear time-invariant systems analysis.
The performance (of the task performed) shows a graceful degradation with noise.
With this motivation we are focusing on issue of efficient sampling, reconstruction, representation
and coding of audio signals. In this research, we are exploring the following sub-topics which could
yield new techniques for information and signal reconstruction with focus on auditory processing like
A. sparse representation (for example compressive sampling) based reconstruction,
B. random sub-Nyquist, event-triggered sampling
C. noise assisted signal processing (example using stochastic resonance)
A Doctoral Thesis ... a body of research that, in a small way, will move a field forward.
Random Processes, Pattern Recognition and Neural Networks,
Time-Frequency Analysis, Adaptive Signal Pro- cessing, Matrix Theory, Digital Signal Compression,
Non-linear Signal Processing, Stochastic Models for Speech Recognition, Digital Image Processing,
Introduction to Neuroscience. The above courses are the few I took from the big list at IISc.