Comparative Study of Classification Algorithms in Sign Language Recognition

Sign language can be used to converse with other people who are HSI. But there is the issue of many people not being able to properly understand due to the lack of people who know sign language. Therefore, image processing and different algorithms with machine learning can be used to recognize signs made by people. Even though sign language recognition has been possible there are many variations of sign language while most only focus on a single type of sign language while majority use ASL. In this article, two types of sign languages have been used to classify and train different types of machine learning models. The images have been preprocessed using background subtraction and then finding the edges using canny edge detection. This articles also goes through the different methods in which the image can be preprocessed. These have been performed using OpenCV library. Two different feature extraction algorithms have been used, namely ORB and SIFT, to perform feature extraction. K-Means Clustering has been used on the features and a histogram is obtained for each sign by using K-Means. Two different datasets, each for ISL and ASL have been split for training and testing in 80:20 ratio and then different machine learning models from sklearn library have been trained using this and the results compared.