Journal of Psychiatry

Design and implementation- a fuzzy classifier for detecting the emotions of autistic children

27^th International Conference on PSYCHIATRY & PSYCHOLOGY HEALTH

June 18-19, 2018 Paris, France

Mahsa Naeeni Davarani and Ali Arian Darestani

Islamic Azad University, Iran

Posters & Accepted Abstracts: J Psychiatry

Abstract:

The study of emotions has always been a matter for philosophers and psychologists. A complete process of face detection involves three steps that include face detection, feature extraction, and cognition. We tried to focus on all three stages as main focus of this paper. To identify facial expressions in eight states (neutral, anger, ridicule, hate, fear, happiness, sadness and surprise), Cohn dataset is used and then the pre-processing is considered to identify the face box. Using the LBP algorithm, the edges of face lines will be determined. The obtained dimensions from the algorithm LBP are reduced by PCA. The simulation results show that by reducing the dimensions, accuracy has considerable improvement in three-layer perceptron neural network. The accuracy rate in the neural network, without PCA is equal to 37.72 due to limited data and abounding features; and by applying PCA, the accuracy will be 86.61 in the three-layer perceptron neural network. The simulation results with the neural-fuzzy network also show that if a more complicated network is used for a linear problem, it will not work efficiently because the accuracy rate of the neural-fuzzy network is equal to 76.51%, which is about 10% lower than the three-layer perceptron neural network. Finally, a system with accuracy of 72.50 will be implemented for classifying autistic children using the healthy children and autistic children's database.