Pre-Temporal Alignment of Sign Language Videos

Sign language is a language which is used by deaf people. They can express their feelings and thoughts by combining gesture, movement of body and face expressions. Spatio-temporal methods are used for Sign Language Recognition (SLR) in previous studies. However, there are some Sign Language Recognition difficulties about misclassification due to temporal differences, alignment different signs or sign comparison. We use a novel temporal alignment method that can be used in combination with many different spatial representation techniques. Our Sign Language Recognition method aims to align given signs and determine key frames before classification step. To find optimal alignment between two videos, we used Dynamic Time Warping algorithm. Our method improves SLR performance on the Bosphorus Sign Toy Dataset.