Facial Feature-Based Human Emotion Detection Using Machine Learning

Abstract

Human emotion detection from video frames is among the most robust and difficult tasks in the field of social communication. Human emotion recognition has attracted the interest of many problem solvers who are working in the artificial intelligence (AI) domain. The emotions on a human face are very subtle and can express their thought process. It also provides a glance at thoughts happening inside the human mind. Emotion detection is the process through which different aspects of human facial expressions are recognised or identified. These, like happiness, anger, sadness, disgust, amazement, fear, neutral and more, are recognised or identified. Primarily, it is found that with the modifications in the emotional phase the body gesture of a person also changes completely. The changes are also visible in body movements, speech, facial expressions, gestures, etc. These parameters or body language are prominent features for automatic emotion detection using machine learning (ML) and deep learning (DL). The applications for human emotion detection based on ML tries to understand the meaning of these body language attributes and implements this knowledge to different available information and datasets. The various advanced algorithms based on ML and DL like principal component analysis (PCA), random forest, support vector machine (SVM), k-nearest neighbor (kNN), etc., are proposed by different researchers. These algorithms can help in the extraction and recognition of facial expressions, voice features, biological signals, body gestures, etc. The accuracy of these algorithms outperforms several datasets of images and videos. The correct ML algorithm helps to achieve accurate results in the detection of human emotion. This chapter presents the benefits and limitations of the numerous methods used for the detection of human emotions. However, the importance of theory has its limitations over the available simulation software. The researchers working in the human detection field are therefore consistently looking for new methods of measurement.