Robust Facial Emotion Recognition using Marine Predators Algorithm with Deep Convolutional Neural Network

Abstract

Facial emotion recognition (FER) is a technology that includes the automatic identification and categorization of human sentiments depending on facial emotions. It leverages deep learning (DL), computer vision (CV), and machine learning (ML) methods to analyze the features of an individual's face, like the place of the mouth, eyes, eyebrows, and complete facial movements for determining their emotional conditions. Popular emotions that FER can recognize comprise surprise, sadness, happiness, fear, disgust, and anger. FER employing DL has an advanced application of artificial intelligence (AI) and deep neural networks (DNNs) that contain training methods to automatically recognize and categorize human expressions based on facial expressions. This approach normally comprises convolutional neural networks (CNNs) or highly complex models namely recurrent neural networks (RNNs) and convolutional RNNs (CRNNs) for analyzing and interpreting complex facial features and dynamics. This study introduces a new Robust Facial Emotion Recognition employing the Marine Predators Algorithm with Deep Learning (RFER-MPADL) approach. The main aim of the RFER-MPADL technique is to detect and categorize different kinds of emotions expressed in facial images. To accomplish this, the RFER-MPADL technique initially applies a bilateral filtering (BF) approach for the preprocessing step. Additionally, the RFER-MPADL technique uses the EfficientNet-B0 method for feature extraction. Moreover, the tuning process of the EfficientNet-B0 method was implemented using the MPA. Finally, the classification of facial emotions can be performed by the use of a deep autoencoder (DAE), in turn augments the overall performance of the RFER-MPADL method. The experimental analysis of the RFER-MPADL methodology is assessed on a standard facial expression dataset. The extensive outcomes exhibited the effectual performance of the RFER-MPADL methodology over other methods.