Classification of ASL Alphabets and Numbers using ORB and Fast with Brief Feature Extraction and Dimensionality reduction techniques

In recent years there has been an exponential growth in the Computer Vision field, especially in the object classification domain. In this paper, various methods are compared to classify the American Sign Language (ASL) with the help of some well-known OpenCV feature descriptors and dimensionality reduction techniques. This project aims to classify signs using traditional machine learning algorithms with the best dimensionality reduction technique to get a quick response from the model and avoid complexities like seen in a CNN model. The proposed methods are compared in terms of maximum feature extraction, dimensionality reduction and accuracy of various classifiers out of which ORB with k-means has the highest accuracy of 97.08% with KNN classifier.