标题：PERFORMANCE EVALUATION OF COMBINED CONSISTENCY-BASED SUBSET EVALUATION AND ARTIFICIAL NEURAL NETWORK FOR RECOGNITION OF DYNAMIC MALAYSIAN SIGN LANGUAGE
期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2017
卷号：95
期号：11
出版社：Journal of Theoretical and Applied
摘要：Dynamic sign language recognition is important to intelligent human-computer interaction technology, but it is also very difficult to deal with, especially when the environment is quite complex. This paper proposes the use of consistency-based Subset Evaluation and Artificial Neural Network (ANN) in order to increase accuracy rate in the recognition of sign language. At the first stage, image acquisition data gets from the Kinect sensor, using skeletal data tracking with eight joint positions. The second stage, the skeletal feature extraction (data X, Y, and Z are taken the value relative to the torso and head; spherical coordinate a conversion process; segmentation of the frame to get the same number of dimensions). The third stage, the selection of feature data using the Consistency-based Subset Evaluation algorithms with the best first search method subset. Then last stage is the classification using Artificial Neural Networks, by variations of nodes in hidden layer. The data samples tested are 15 dynamic signs taken from the dynamic of Malaysian Sign Language (MySL). The results of the experimental show that our system can recognize with accuracy of 93.67%. The feature selection can contribute toward the improvement in the accuracy rate of sign language recognition data
