Design of an adaptive winner takes all circuit explaining features of binocular rivalry in visual brain

Abstract

The paper presents an adaptive WTA circuit where the firing of a neuron depends on input stimuli and the circuit self-adapts which in turn allow the loosing neuron to win. The CMOS based design have been simulated at 45 nm CMOS process which occupies less chip area (of about 12 × 25 μm2) and consume low power (about 124 nW). The design is stable as temperature coefficient is about 11.5 μV/°C and noise figure is about 11.6 nV/Hz. The self-adaption feature depends on the value of capacitor ratio which is equal to 50 fF. A visual brain feature called Binocular Rivalry (BR) has been demonstrated and explained using simulated results of proposed circuit. Three features of visual brain; superimposition of images obtained from two eyes; binocular rivalry when different images are presented to eyes: suppression and dominance of images, and neural adaption, have been explained using circuit simulation. The real time images processed using image processing illustrating BR features has also been implemented to circuit for real time operations. The time taken for neural adaption which causes switching of images in human brain when distinguish images are provided to eyes is about 0.2s, as demonstrated in circuit simulations using BR images. The proposed design can be implanted in machines (humanly robots) with which it can see, learn and self-adapt like a human visual brain. The circuit and the concept can also be implemented in devices used in assistive technology for helping visually disabled or less able persons. The Papers proposes a novel idea of CMOS based hardware for implementing BR features in machines.