Title "A moving object recognition by spiking neurons" introduction - The detection of moving objects from video frames - Identifying moving objects from a video sequence - identifying, tracking and classifying moving visual ovjects On one hand it is a very practial ... Let's name a few. traffic monitoring, video surveillance military applic soccer --- eg help devising a strategy commonly based on background subtraction at the same time it's interesting to approach this issue from cognitive neuro science, that is, how human or animal identify/recognize/detect moving objects with visual cortex Wu " ... spiking neural network which is used to simulate the visual cortex ... N1 or Neuron N2 fires. " e.g. we can connect input output and programmed quite easily to establish such input-output relation. -------------------------------------------------------------------------------------------------- The goal of the project is identifying moving objects from a vedeo sequence. On one hand, we can apply it practical issues such as traffic monitoring (see e.g. ... ...), video surveillance (see e.g. ...), military applications, football scnario, sign language or lip-movement recognition (see ...) We see surveillance video camera everywhere nowadays. Detection of only movement objects helps us reduce memory needed. Or movement detection from soccer video helps to extract good or bad strategy in football game. Furthermore, recognize lip-movement and translate into text would be great application. On the other hand, it's interesting to approach this issue from cognitive neuro science, that is, how human or animal identify/recognize/detect moving objects with visual cortex ... -------------------------------------------------------------------------------------------------- benchmark author wrote conductunce IF is very close to ... while Izhikevich ... simple IF has detect a moving object from an image sequence in which moving static image as a background created at random "The network consists of N = 1000 neurons with the first Ne = 800 of excitatory RS type, and the remaining Ni=200 of inhibitory FS type (Izhikevich, 2003)." => Izhikevich specifically uses RS type for excitatory neurons and FS type for inhibitory neuron. but how about another combination? "The ratio of excitatory to inhibitory cells is 4 to 1, as in themammalian neocortex. Each excitatory neuron is connected to M=100 random neurons, so that the probability of connection is M/N = 0.1, again as in the neocortex." as he put it, we also follow this ratio Future work As the author pointed it out the delay is not adactive. We do not model modifiable delays (Huning, Glunder, & Palm, 1998; Eurich, Pawelzik, Ernst, Cowan, & Milton, 1999) or transmission failures (Senn, Schneider, & Ruff, 2002). Implement let me paraphrases Wu et al. ... input(x y) is connected intermediate n1(x y) connected through "excitatory synapse without delay & inhibitory synapse with delay" and n2(zxy) is connected through excitatory with delay and inhibitory without delay both excitatory and inhibitory synapses are adjusted so that neuro N1 does not fire when g(x,y,t)=g(x y t-dt) Let S_{xy}(t) represents spike train generated by Neuron N_{xy} in output layer. The firing rate for Neuron N_{xy} is calculated by the following expression. r_{xy}(t)=(1/T)\sum_t^{t+T} S_{xy}(t) Plotting r_{xy}(t) as a grey image, white areas indicate neuron groups with high firing rate. Drawing the outside boundaries of firing neuron groups, boundaries of moving objects are extracted. The edges of firing neuron groups are used to determine the boundaries of the moving objects. Let SNr(x, y, t) represent current from receptor Nr(x, y). As for rather than wu et al in which ... are not clearly specified. therefore we follows following Florian, R. V. (2005) input The activations were normalized between 0 and 1. The input neurons fired Poisson spike trains, with a firing rate proportional to the activation, between 0 and 50 Hz. output ================================================================================================== Reference Q. Wu, T.M. McGinnity, L. Maguire1, J. Cai, and G. D. Valderrama-Gonzalez (2008) "Motion Detection Using Spiking Neural Network Model." Proceedings of the 4th international conference on Intelligent Computing: Advanced Intelligent Computing Theories and Applications - with Aspects of Artificial Intelligence. Lecture Notes In Artificial Intelligence Vol. 5227. pp. 76--83. => done Izhikevich, E. M. (2003). Simple model of spiking neurons. IEEE Transactions on Neural Networks, 14, 1569.1572. IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL. 15, NO. 5, SEPTEMBER 2004 1063 Which Model to Use for Cortical Spiking Neurons? Eugene M. Izhikevich Florian, R. V. (2007) gReinforcement learning through modulation of spike-timing-dependent synaptic plasticity.h Neural Computing, Vol. 19, No. 6. pp. 1468-1502. Florian, R. V. (2005) hA reinforcement learning algorithm for spiking neural networks.h Proceedings of the Seventh International Symposium on Symbolic and Numeric Algorithms for Scientific Computing, pp. 299.306.