Advanced Deep Convolution Based Jellyfish VGG-19 Model for Face Emotion Recognition

Abstract

For many applications, facial emotion recognition (FER) is an essential yet unsolved procedure. In the past, artificial intelligence methods like convolutional neural networks have typically been used to recognize emotions. However, in terms of complexity and processing time, this method is quite costly. An optimization-based deep convolution network that uses attention-based Densenet-264 for feature extraction is presented in order to address this issue. In the first step, the images are pre-processed using image resizing and equalized joint histogram-based contrast enhancement (Eq-JH-CE) to enhance the image quality. Next, an enhanced attention-based DenseNet-264 architecture is developed for feature extraction, which helps improve classification accuracy. Finally, the extracted features are used by the Advanced Deep Convolutional based Jellyfish VGG-19 model (DeepCon_JVGG-19) for classifying face emotions like angry, disgust, fear, happy, neutral, sad, and surprise. Here, Jellyfish Optimization is used to fine-tune the optimal parameters and increase the performance of the classified model. The Python tool is used for implementation. The JAFFE and FER-2013 are used to test the proposed model performance. The experimental analysis proves the strength of the proposed study by attaining 98.5% accuracy.