ASIC Implementation for Improved Character Recognition and Classification using SNN Model

The third generation of spiking neural networks raises the level of biological realism by using individual spikes.So instead of using rate coding, these neurons use pulse coding mechanisms where neurons receive and do send out individual pulses, allowing multiplexing of information. This work depicts how Spiking neural network model is used for character recognition and classification. Here, we adapt to the technique of using ASIC for large scale simulations of the Izhikevich model and use RTL Clock gating approach for reducing the dynamic power. Here the focus is on how power consumption and system cost can be reduced for large production run. The full custom biologically plausible spiking neural network model is implemented on ASIC with 90 nm Process. The Izhikevich spiking neuron model is best suited for large scale cortical simulations due to its accuracy, efficiency, power and simulation time. The classification efficiency of SNN based on MATLAB simulations is demonstrated in this work by its ability to classify the 27 characters correctly out of 30 noisy character images presented. The ASIC realizing the English character classification and recognition dissipates power of 2.8 mW and an area of 120312 μm2. This work brings about the application of using networks of these spiking neurons for character recognition and their suitability for custom realization with reduced power consumption.