Time varying images sparse coding network model based on simple cell in visual cortex

Study in neurobiology has shown that properties of spatio-temporal structure of receptive fields in mammalian visual perceptual mechanism emerge from sparse presentations towards continuous time varying stimuli. According to this theory, we present a sparse coding network model. Time varying images sequences as inputs are used to form receptive fields both in space and time domain so as to exhibit comprehensive properties of receptive fields. Bilateral neural network with feedback provides robustness to noise. Orthogonal basis and self inhibition function strengthens sparseness of the code. Our model´s plausibility in neuroscience view and feasibility in practical computation will also be discussed.