Stereo-vision calibration by multi-layer perceptrons of an artificial neural network

Abstract

The aim of this study is to restore some vision to the visually impaired individual via non-invasive digital stimulation. The objective is attained by transmitting processed information to the vision cortex, via non-invasive Trans-cranial Magnetic Stimulation (TMS) technique so as to excite the elementary unit of the nervous system. The feeding of the vision neural system is done through an Artificial Vision Neural Network (AVNN) of interconnected cells that generate images which help the vision impaired individual to recognize objects. The AVNN adapts the processed information to a form suitable for transmission. The said information is wirelessly sent by an array of electrodes to the brain; that is, a neuron to neuron communication is applied. This process is easy to apply since the function and the structure of the AVNN are similar to those of the neurons found in human's vision system, and which can be replaced by a well learned network. The AVNN network is fed with the 3D image provided by a digital stereo vision system, which by itself is an auto updated stereovision neural network. In this study, the calibration phase by neural network is achieved.