Feature binding neural model based on a coupled oscillator system

We propose a neural network model for feature binding based on phase locking of a coupled oscillator system. The proposed model can represent positions of visual stimulus and detect binding errors in the positions of visual stimuli. It is assumed that binding and segmentation is achieved by oscillation. Binding among different features is achieved by oscillating in phase. Segmentation among attributes in a feature is achieved by oscillating in antiphase. By computer simulations, we show that the proposed model can reproduce a process of binding features with position information based on phase locking. In our model, the phenomenon of illusory conjunction can be considered as the transition state to phase locking. Simulations show that phase locking appears in the early stage when the number of stimuli is small. It is considered that this behavior corresponds to the decrease in probability for illusory conjunction in human vision.