| Summary: | 碩士 === 國立東華大學 === 資訊工程學系 === 104 === In recent years, people pay more and more attention to the protection of their privacy. On many personal electronic appliances, different types of electronic mechanisms for person identification are very common. The most common one could be the password verification systems. Usually, users can enter the passwords by key typing, handwriting or hand drawing. However, there is still a risk of privacy leakage that the entered passwords can be exposed to malicious snoopers. To reduce the risk, this paper aims to design a lip reading system for numerals so that users can enter the passwords with voiceless speaking. In the study, we compare the recognition performance of many different features, including the Local Binary Pattern (LBP) and Histogram of Oriented Gradients (HOG), using the Dynamic Time Warping (DTW) method for recognition. By experimenting on the public CUAVE database (English numbers) and a prepared Chinese numbers database, the best recognition rate we have achieved is about 85.5%.
|
|---|
