FPGA based Structural Radial Basis Function Neural Network with Hybrid Optimization for Neural Activity

Emotion Recognition is one of the prerequisites for Human to Machine (H2M) cognitive communication. H2M communication using Electro-Encephalogram (EEG) based cognitive interfaces lowers overall reaction latency and improves decision-making throughput of cyber-physical systems. This work presents an emotion classification technique that is suitable for deployment in edge based IoT devices with sufficiently high accuracy and specificity compared to state-of-the-art architectures. Proposed technique has been verified on a real time PYNQ Z2 based hardware accelerator platform.