A.V. Williams Bldg., Rm. 2460
For More Information:
301 405 3596
"Changing your mind in a cocktail party scene:
How does studying the neural bases of attention and learning in the auditory cortex inspire algorithms for analysis and recognition of speech and music?"
Prof. Shihab Shamma
Anyone who has walked into a crowded reverberant hall, with music blaring in the background, will recall the initial impression of the sound as a loud and undifferentiated noise. In short order, however, different sound streams begin to emerge as one attends to a few speakers, listen to the melody from the band, or even to one instrument in it. Humans perform this remarkable feat effortlessly, but so do many animals too. Natural environments are often extremely cluttered and disorienting, and hence animals have developed abilities to navigate their complex auditory scenes in order to mate, feed their newborns, and avoid predators. For instance, Penguin parents locate their nest by the cries of a chick in the midst of a millions-strong colony of screaming birds often situated on a flat terrain without any physical landmarks. Frogs also locate singing mates in the midst of a cacophony of calls. To accomplish these feats, it is likely that similar mechanisms have evolved in many animals which include a mix of "bottom-up" automatic processes with complex "top-down" behaviors involving attention, expectation, learning, and memory.
Our research attempts to sort out these parallel processes and explore the strength and dynamics of their interactions. For example, in the "bottom-up" category of mechanisms, animal brains are largely hard-wired to respond and perceive sound so as to highlight the features important for the survival of the species. These might include sensitivity to specific frequency ranges or repetition rates, or an inherited ability to readily learn to recognize con-specific sounds early in life, such as speech for human infants, songs for many birds, and mating calls in frogs. In mammals, the key machinery to perform this sound analysis reside in the auditory area of the cortex, where many neurons have been found to respond selectively to specific spectral and dynamic features, a selectivity which is often referred to as the receptive field of the neurons. It has been known for sometime that different regions in the auditory cortex exhibit an ordered diversity of receptive fields that emerges automatically during development. It has also been confirmed that this organization is not fixed forever in the adult, but rather is plastic and can be modified in response to injury or following long-term training which can enhance response sensitivity of some neurons and hence re-allocate the proportion of neurons responding to different aspects of the sound.
Dr. Shamma's research deals with issues in computational neuroscience, euromorphic engineering, and the development of microsensor systems for experimental research and neural prostheses. Primary focus has been on studying the computational principles underlying the processing and recognition of complex sounds (speech and music) in the auditory system, and the relationship between auditory and visual processing. Signal processing algorithms inspired by data from neurophysiological and psychoacoustical experiments are being developed and applied in a variety of systems such as speech and voice recognition and diagnostics in industrial manufacturing. Other research interests included (at various times) the development of photolithographic microelectrode arrays for recording and stimulation of neural signals, aVLSI implementations of auditory processing algorithms, and development of robotic systems for the detection and tracking of multiple sound sources.
This Event is For: Clark School