Abstract: Human emotion recognition has significant part in our day to day lives. Objective of the study is to develop and implement a system capable of analyzing, predicting, and classifying emotions in real-time using Convolutional Neural Network (CNN) algorithm, with the assistance of the OpenCV library. The proposed approach enables the classification of various emotions, including anger, disgusted, fear, happy, neutral, sad, and surprised, based on feature extraction. FER2013 dataset is utilized for performance evaluation, and pre-processing techniques such as facial landmark detection are employed during training and testing.
This dataset is utilized for training and testing purposes, as it is understood that while one-third of communication is conveyed verbally, the remaining two-thirds are conveyed through non-linguistic. means. Although there are existing emotion recognition systems, in real-life scenarios, consider the example of mental hospitals where this technology provides medical professionals with insights into patients' emotional states. By leveraging this technology, medical professionals can offer improved care and potentially enhance outcomes. Facial expression recognition remains a challenging problem in computer vision, as images of the same person in different expressions can change in brightness, background, and position.
Keywords: Emotion recognition, Convolutional Neural Network (CNN), OpenCV, Pre-processing.
|
DOI:
10.17148/IJARCCE.2023.125241
