An Efficient Lip-reading Method Using K-nearest Neighbor Algorithm

Abstract

Many studies have been carried out on lip reading, most of those works are based on color images, while some essential features might not be obtained, like inner lip information. In this paper, RGB-D camera will be introduced for improving the recognition rate of lip reading. We try to complete lip reading through using only gray-scale images. Thirteen groups of words are given, and we present eight features for classification. Volunteers are asked to sit in the front of RGB-D camera. For each word we select 15 frames. K-nearest neighbor algorithm (KNN) is used to select the same words between different volunteers.

DOI: http://dx.doi.org/10.11591/telkomnika.v13i1.6872