Evaluation of Autism Classification Using Machine Learning Techniques

Autism is a development disorder that impacts the social interaction, cognitive abilities and attention of children. The child's caretaker can identify the autistic behavior visually when a child displays repetitive behaviors, limited eye contacts, limited facial expressions and no signs of imitation. The proposed work uses GLCM and SVM for classifying autism and normal controls. GLCM is used for feature extraction from thermal face images and SVM is used for the classification process. The Feature extraction is done to extract the features like contrast, correlation, cluster prominence, cluster shade, dissimilarity, energy, entropy, homogeneity, maximum probability, a sum of squares, variance and autocorrelations. Machine learning algorithm GLCM and SVM are trained and validated with 40 thermal images of different subjects. Through various experiments, the classification sensitivity and specificity are found to be 66.58% and 91% respectively. The accuracy level and normal subjects are 86.73 and autism subjects 85.30 respectively by using GLCM and SVM.