Sign Language Classification Using Search Algorithms

We concentrate on the problem of feature selection on sign language classification. The goal is to maximize the classification accuracy using a proper subset of features out of totally 22 given features. In this work, we employ two search methods: Hill Climbing approach and Random Walk approach, to select the features. We claim that both algorithms are easy-implemented but reasonable and efficient. We illustrate the power of our methods by successfully achieving a relatively high classification accuracy on a real dataset.