Recognition Of Global Features In Rats’ Primary Visual Cortex Based On Manifold

After hundreds of millions of years’ evolution, animals’ visual system caneffortlessly complete the task of detecting and identifying special targets. Its speedand efficiency are beyond comparison with those of artificial computer vision andimage processing methods. In visual system, primary visual cortex is the hub of thetwo information pathways; the vast majority of information from the retina isprocessed in the primary visual cortex and passed to the next level. Therefore, studythe response characteristics in the primary visual cortex under the stimuli of differentglobal features and use these responses to recognize different global features are ofvital significance in revealing how the visual system can recognize specific objectsquickly and accurately.In this thesis, rat was chosen as a model animal, the response characteristics inthe primary visual cortex were used to study the recognition of global features. Themain research and findings are listed as follows:(1) The receptive field of neurons in primary visual cortex was tested.According to the different response characteristics of neurons in primary visualcortex, the checkerboard stimulation pattern was combined with spike-triggeredaverage to determine the receptive field. The experimental results show that: in thiscase, the receptive field of neurons near the position where electrodes are implantedis at left of the center of screen. The total firing number of neurons is measuredunder two circumstances: stimuli appear in the receptive field position and in theother non-receptive field position. The former number is much larger than the latter.(2) The response time interval of neurons to global features was determined andverified.According to neurons in primary visual cortex exhibit a high degree ofsimilarity to the visual stimuli with the same global features but different localfeatures in the beginning of the response time, the same global features but differentlocal features were tested in the in the receptive field. The firing rates were used to determine the effective response time interval to global features. By adding up theeffective response time interval of multiple neurons in primary visual cortex underthe stimuli of global features, a rule is derived: the effective response time interval ofthe neurons in rats’ primary visual cortex is between100ms and350ms. After usingeffective response time interval, classification accuracy has been raised greatly.(3) The response characteristics were measured and which responsecharacteristics of spikes carry more information of global features was determined.In the effective response time interval, the response characteristics includingaverage firing rate and firing intervals of single neurons were measured. Whichresponse characteristics of spikes carry more information of global features wasdetermined by the last classification accuracy. The spike firing rate and inter spikeinterval are carrying information associated with the global features, but in contrast,the composite characteristics carry more information associated with the globalfeatures.(4) Different manifold approaches were used to integrate the characteristics.Two kinds of manifold approaches including principal component analysis andisomap were used to integrate the characteristics, support vector machines were usedto classify the characteristics integrated. Overall, under the same circumstances, theclassifying precision achieved by using isomap is higher than those achieved byusing primary component analysis.