Improving Facial Expression Recognition with Deep Learning for Faces of Different Ages

Abstract

Many times automatic human emotional detection plays a vital role in health care and law enforcement agencies. To achieve this, it is necessary to have a robust hardware resource that can capture images under harsh conditions; numerous components already exist in the industry for this purpose. However, the greatest challenge lies in developing an advanced algorithm to process these images and extract the desired emotions. Many image processing techniques, such as PCA and transform techniques, are available. But these technologies aren't providing satisfactory results in harsh conditions. So, the evolution of deep learning models plays a crucial role in the identification of faces and, thereby, emotions. Keeping this fact in mind, designed model employ a deep learning model to enhance the image channels, which can be then use to efficiently train a large dataset. Therefore, the designed model utilizes a channel boost convolution neural network to achieve optimal accuracy in identifying facial emotions across a broad age range. The results show that the designed model outperforms other emotion recognition models, achieving a root mean square of 0.2948 and an accuracy of 98.34% on a variety of age facial images.