A Machine Learning Model for Kids’ Behavior Analysis from Facial Emotions using Principal Component Analysis

Identification of the emotional state of humans, especially kids’, is a very complex activity. Different types of emotions contribute to the behavior of kids. There are various methods to recognize the emotional state like verbal communication, non-verbal gestures like movement of hands, voice tone and facial expressions. Among these, recognition of the facial expressions is the most widely used method to characterize human emotions further to predict human behavior. In this work, a machine learning model is proposed to recognize the emotional state of the kids’, i.e., toddlers and preschoolers. Proposed model is based on PCA technique and MLP classifier. Data set is pre-processed using gradient filtering and extracted features are optimized using PSO. Training data used in this work, comprise of 273 facial images of the kids in the age group of 2 to 5 years. Dataset belonged to four facial expressions, i.e., happy, sad, neutral and thoughtful. Proposed model gave better results in comparison to two existing model with an accuracy of 95.63%. The proposed model can further be enhanced for emotion recognition and behavior analysis of mentally retarded kids.