Spiking neural network for visual pattern recognition

Most of visual pattern recognition algorithms try to emulate the mechanism of visual pathway within the human brain. Regarding of classic face recognition task, by using the spatiotemporal information extracted from Spiking neural network (SNN), batch learning rule and on-line learning rule stand out from their competitors. However, the former one simply considers the average pattern within the class, and the latter one just relies on the nearest relevant single pattern. In this paper, a novel learning rule and its SNN framework has been proposed. It considers all relevant patterns in the local domain around the undetermined sample rather than just nearest relevant single pattern. Experimental results show the proposed learning rule and its SNN framework obtains satisfactory testing results under the ORL face database.