各位老师：

您好！美国Duke  University的李征博士将来北师大讲座，欢迎感兴趣的老师和同学参加。讲座信息如下：

主题：Computational aspects  of recording and decoding motor neural activity from Rhesus cortex

时间：2012年7月5日上午9:00

地点：小楼三层大会议室

主讲人：李征博士

主讲人简介：

Zheng  Li received a B.S. in computer science and mathematics from Purdue University in  2004 and a Ph.D. in computer science from Duke University in 2010. He is  currently a postdoctoral associate at Duke University in the lab of Miguel  Nicolelis. His research interests are invasive brain-computer interfaces for  prosthetics and other applications

内容摘要：Brain-computer interfaces for  controlling prosthetic devices translate (decode) neural activity to movement  commands. These systems have the potential to restore movement to people  suffering from paralysis. In this talk, I will present our work on the  computational side of this multi-disciplinary field. First, I will discuss a  decoding algorithm which uses a model of neural tuning which includes modulation  for movement speed as well as direction. This decoding algorithm, the unscented  Kalman filter, facilitated brain-control of a computer cursor by a Rhesus monkey  with higher accuracy than previous decoders which did not include this speed  dependency. Another important consideration when decoding is adapting to changes  in neural tuning over time. I will present a method for updating the model  parameters of a decoder automatically, without user intervention. Our method  uses a Bayesian update process that intelligently combines previous model  parameters with new model parameters fitted from incoming data. Using this  method, a Rhesus monkey retained control of a cursor in experimental sessions  spanning almost a month. Lastly, I will present our recent work on recording  neural activity from many neurons simultaneously. We have developed a networked  recording system consisting of multiple desktop computers with analog signal  acquisition cards and running custom software. The system is designed to allow  scaling-up to recordings from thousands of electrodes. The analog signal  acquisition cards capture amplified voltage traces from electrodes while our  custom software performs spike detection, spike sorting, and delivery of spike  information across the network. I will conclude with a discussion of topics for  future research.

    此致

敬礼

认知神经科学与学习国家重点实验室

2012年6月26日