Classification of Turkish words by using lip motion features.

Information obtained from Lip Reading in addition to voice data is used in voice therapy of hearing-impaired persons. Although there are many studies on the applications of Lip Reading, no study exists on the recognition of Turkish Words using MS Kinect camera, which has a built-in integrated infrared sensor that measures the depth information. The aim of this project is to construct a data set on frequently used Turkish words containing depth information, investigating the best lip reading classification method on this data set. For classification, techniques such as Artificial Neural Networks, KNN, and Dynamic Time Warping has been used. Furthermore, a complementary software was developed, which improves the lip imitation skills of hearing-impaired children and people who need speech therapy with the aid of an instructor by comparing two pronunciations. The developed software and the generated data set will be used as educational materials for hearing-impaired children.