Feature Extraction Based on Cellular Particle Swarm Optimization Algorithm for American Sign Language Images

Abstract

Image Recognition is becoming a critical task and many problem-solving systems and approaches for image detection, analysis and classification are introduced by many modern researcher. The techniques should be user friendly and ease to interpret for both normal people and special people, it they should be analysed. The sign language act as a mode to transfer and exchange the message, information, knowledge and ideas from deaf to common people. The gaining information and responses to the pattern or gesture is called as sign. Sign Language is only mode to communication between the hearing-impaired people and common people. In this method each individual gesture is called as sign. In this system the American Sign Language recognition system which is used for visually impaired people to access and educate the special children’s. American Sign Language (ASL) images of a special people is collected with some constraints are taken as the dataset. The objective of the research work is done with three different Phase. Phase1 is pre-processing, Phase 2 is segmentation and Phase 3 is feature extraction for static hand gestures with the maximum possible accuracy rate. A novel approach for the proposed system is analysed with its structural feature extraction and compared with its parameters. The novel feature extraction is made with 8 different structural features they are Bounding Box, Area, Perimeter, Centroid, Roundness, EquiDiameter, Number of Boundaries and Angle. F-Measure, Recall and Precision based on this Accuracy, Sensitivity and Specificity are measured for features are extracted images. The accuracy of the new Phase has been found significantly higher.