A Novel and Enhanced Facial Electromyogram Based Human Emotion Recognition Using Convolution Neural Network Model with Multirate Signal Processing Features

摘要

In recent era, emotion recognition plays a vital role in human–machine interaction. Service robots identify the significant aspects of human behavior by analyzing the emotions of human beings and make use of these emotions to communicate with humans which make the human–robotinteraction faster. Human emotional states are different and they are exhibited instinctively which are not easily recognized by the self-service robots. This paper attempts to develop a novel and enhanced multimodal emotion recognition system which could easily identify the six emotions namelyanger, disgust, fear, happy, neutral and sad of an individual. The developed FEMG model could be interfaced with any service robot so that the human–machine interaction could be made faster. Facial Electromyogram (FEMG) signals are acquired from 20 subjects. Multirate signal processingfeatures namely Multidecimate, Multidownsample and Upfirdn are applied to the FEMG signals. A deep learning technique namely, Convolutional Neural Network (CNN) model is applied to the feature extracted signals and the highest classification accuracy obtained was 99.79%. Also the developedFEMG based CNN model also reduces the training time employed in the experiments.