Statistical features based selfie sign language recognition system (SSLRS) with ANN classifier.

This work is to bring mobile based sign language recognition system into real time. Selfie sign videos are captured with smart phone front camera. Morphological gradients along with Sobel edge operators are used to extract hand contour from each sign video frame. Discrete Cosine Transform (DCT) of hand contour is optimized by principle Component Analysis (PCA) to reduce the execution time. The four statistical features such as mean, skewness, standard deviation and kurtosis are calculated for the optimized hand contour DCT. The feature vector formed with these four statistical features is used for sign classification using Artificial Neural Networks (ANN) classifier. The performance of SSLRS is evaluated with the Word Matching Score (WMS). [ABSTRACT FROM AUTHOR]