Emotion Classification of Facial Electromyogram Signals Based on Neural Networks

Signal processing using statistical methods is a remarkable area in the field of practical applications in Mathematics to investigate and mine the most essential facts from the signals. It removes the unwanted noise, deals with their statistical properties like standard deviation, covariance, median, average etc. Being a versatile feature extraction method, it is also used in different areas such as natural language. processing, bio signal processing, sonar. In this work, we have examined the facial electromyography signals (FEMG) by applying the statistical features in order to extract the necessary features for categorizing the six emotional conditions namely happy fear, neutral, sad, disgust and anger. 20 subjects took part in this FEMG study. The statistical features namely kurtosis, skewness, moment, range, median absolute deviation and mean were used in this work to draw the features. In order to classify the emotive conditions, the four neural network models namely Cascade Neural Network model, Elman Neural Network model, Layered Recurrent Neural Network model and Feed Forward Neural Network model were modelled accordingly. Outcomes of this work indicate the highest