TITLE: Visual Signal Processing and Biological Pattern Generation SPEAKER: Zhijun Yang ABSTRACT The human brain is probably the most complicated system in the world which is composed of some 100 billion neurons with an average of 1000 times of that amount of possibly random interconnections. Neurons communicate with each other by generating and transferring spikes. In mathematics many neuron models have been proposed to mimic the spiking mechanism in different areas of the cerebral cortex. In the primary visual cortex, we used a simple type of integrate-and-fire neuron to build a visual model to recover the lost depth information from the 2D optical flow field. This model also incorporates a popular learning algorithm, namely spike-timing dependent plasticity (STDP), which was discovered just a decade ago. For the pattern generation I will introduce a graph dynamics which is followed by a discrete-time neural network implementation of the graph dynamics. It is shown that this model is able to generate arbitrary rhythmic patterns suitable for driving the legged locomotion. An analog VLSI circuit implementation of a similar continuous-time model of gait pattern generation will also be discussed.