A scene recognition model with using synchronization among non-liner oscillators was proposed by Konig. This model, based on physiological evidence of oscillatory responses in visual cortex, reproduced neuron activities well. However it was impossible to extend its handling information category to both gray levels and high number of pixel, such as HDTV, by the reason of oscillator's inherent character. To overcome these situations, non-linear YANO oscillator that is endowed with wide frequency rage over several hundreds of times is employed to the Konig model. From some computer simulation results, it is clarified that the proposed model should employ newly devised desynchronizing-connection architecture to eliminate alias-correlation that is caused by false pull-in phenomena between oscillators.
抄録全体を表示