CNN-LSTM based Facial Expression Detection System

In a variety of applications, including psychology, virtual reality, and surveillance mental health, facial emotion analysis is a crucial part of human-computer interaction. Using a sizable real-world dataset, this study compares two models: The Viola Jones Haar-Cascade Algorithm and the Convolutional Neural Network (CNN) fusion with Long ShortTerm Memory (LSTM) model. The aim is to correctly identify the six basic emotions-neutral, happy, sad, angry, fear and surprise-as they manifest in a range of face images. To guarantee variety in terms of age, gender, and ethnicity, we start by picking and preparing a large dataset that includes various real-life face expressions. In the suggested methods, CNN is utilised as the main feature extractor to extract spatial information from face pictures and Long Short-Term Memory layer is then incorporated to the CNN-LSTM model to specify the temporal correlations, and the Haar-Cascade classifier is used to categorize the emotions in the Viola Jones model. These techniques, together improves the model’s ability to recognize intricate emotions by enabling it to quickly learn both stationary and moving data.