An FPGA-Based Implementation of Emotion Recognition Using EEG Signals

An electroencephalogram is a machine that uses small metal discs (electrodes) placed on the scalp to detect all the electrical activity in the human brain. Electric impulses connect the brain cells and are active at all times and even while we are sleeping. This activity appears as wavy lines on the EEG recording. The preprocess function filters data to a frequency range of 0 to 75 Hz. It creates a new matrix with a sampling rate of 200Hz and a range of 0 to 75Hz. The Low pass filter of Finite Impulse Response was utilized. Because bandpass would make the EEG data unstable after processing. Each EEG pre-processed signal has output, completing the feature extraction. Principal Component Analysis, or PCA, is used in the feature reduction phase. PCA is a statistical process that turns around a correlated set of features into mutually uncorrelated features, or principal components, using singular value decomposition. Principal Components Analysis: (1) Mean normalization of features (2) Covariance Matrix (3) Eigen Vectors (4) Reduced features or principal components. The preceding step's PCs will be passed into the SVM classifier for emotion output. A VHDL code & testbench for a 2*2 matrix was written and the waveform, RTL schematic was obtained on Xilinx 14.5. For the FPGA implementation, the Simulink Model was designed and the eigen values were computed using a system generator.