IAI Colloquia Series: Behtash Babadi, "Signal Processing for High-Dimensional Dynamic Neural Data"
Wednesday, November 5, 2014
1146 AV Williams Bldg
Intelligent Automation, Inc. Colloquia Series
Robust and Scalable Signal Processing for High-Dimensional Dynamic Complex Neural Data
Electrical and Computer Engineering
Affiliate Faculty, Institute for Systems Research
Thanks to the advances in data acquisition technology, the process of data collection from the nervous systems of animals and humans has been substantially facilitated, resulting in abundant pools of data under various modalities and conditions. These data are typically high-dimensional, dynamic and complex, as they may span hours of brain activity recorded by tens to hundreds of sensors in various spatiotemporal scales. Therefore, modern-day neural data require carefully tailored signal processing techniques in order to be effective in deciphering the many remaining mysteries of the brain.
Converging lines of evidence in theoretical and experimental neuroscience suggest that brain activity is a high-dimensional distributed spatiotemporal process emerging from underlying dynamic sparse structures. In this talk, we will explore how in various scenarios of interest the hypotheses of dynamicity and sparsity can be exploited in a principled way to address several outstanding challenges faced by existing neural signal processing techniques including loss of spatiotemporal resolution due to measurement limitations, loss of spectrotemporal resolution due to time-averaging and windowing, and real-time processing.
In particular, we will present recent and ongoing research in MEG source localization, spectrotemporal analysis of EEG and spiking data, dynamic decoding of auditory attentional modulation from MEG, and adaptive neuronal system identification. Apart from their theoretical significance, these results may bring new insights into sophisticated functional processes of the brain such as auditory processing and loss and recovery of consciousness under general anesthesia.