Workshop 5: Recent advances in Machine Intelligence
Topic 1: Face Recognition from Low-resolution to High-resolution
A lot of research on face recognition has been conducted over the past two decades or more. Various face recognition methods have been proposed, but investigations are still underway to tackle different problems and challenges for face recognition. The existing algorithms can only solve some of the problems, and their performances degrades in real-world applications. In this talk, we will first discuss the performances of face recognition techniques on face images at different resolution. To perform face recognition, image features from a query image are first extracted and then matched to those features in a gallery set. The amount of information and the effectiveness of the features used will determine the recognition performance. To improve the performance, we will present a face recognition approach using information about face images at higher and lower resolutions, which can enhance the information content of the features that are extracted and combined at different resolutions. As the features from different resolutions should closely correlate with each other, we employ the cascaded generalized canonical correlation analysis (GCCA) to fuse the information to form a single feature vector for face recognition. To further improve the performance and efficiency, “Gabor-Feature Hallucination” is used to predict the high-resolution (HR) Gabor features from the Gabor features of a face image directly by using local linear regression. We also extend the algorithm to low-resolution (LR) face recognition, in which the medium-resolution (MR) and HR Gabor features of a LR input image are estimated directly. The LR Gabor features and the predicted MR and HR Gabor features are then fused using GCCA for LR face recognition. This approach can avoid having to perform the interpolation/super-resolution of face images and having to extract HR Gabor features.
For recognition of HR face images, we will show that pore-scale facial features can be explored when the resolution of faces is greater than 700×600 pixels. We will describe the use of the facial features for recognition under conditions of different facial expressions, lighting, poses and captured times. We will also present the minimum area in face images that can retain a high recognition level. Experiment results indicate that the facial pores can be used as a new biometric for recognition, even distinguishing between identical twins.
Topic 2: Speaker Verification under the Big Data Environment
The recent increase in identity theft has drawn the public attention to the potential risk of using remote services. In fact, frauds and identity theft are on the rise in Hong Kong and mainland China because of the popularity of using mobile phones to access remote services. Traditionally, authentication of these services relies on usernames and passwords. This authentication method, however, is no longer adequate for safeguarding the security of remote services. The recent advance in speaker recognition technology has demonstrated that voice biometrics could be an important part of the authentication process. For example, using speaker recognition technologies, the identity of a caller can be verified in a live call and financial transactions can be stopped if fraud is detected. Voice biometrics can also help call centers to reduce the risk of leaking customers’ information caused by social engineering fraudulence.
This talk will introduce the principles behind current state-of-the-art speaker recognition systems. It will cover various components and processing stages of a speaker recognition system, including feature extraction, i-vector extraction, probabilistic linear discriminant analysis, and deep neural networks. Then, the potential problems of the current methods when the data volume increases will be discussed. Finally, the talk will explain how parallel processing technologies such as MapReduce and Hadoop can be applied to overcome the problems in large-scale biometric applications.
About ITS Workshop
Jointly organized by the Department of Electronic and Information Engineering, The Hong Kong Polytechnic University (PolyU EIE), Hong Kong Applied Science and Technology Research Institute (ASTRI), and Hong Kong Science and Technology Park (HKSTP), the Innovative Technology Series (ITS) seeks to keep the industry and the community abreast of the current and emerging technologies in electronic and information engineering. The ITS will be delivered in a series of 12 structural professional training workshops covering different topics of current and upcoming technologies as well as the latest research results. Presenters of the training workshops will be the professors of PolyU EIE and research project leaders of ASTRI. Overseas experts may also join the training workshops to share their experience with local industry and community.
Online registration: http://eie.polyu.edu.hk/event/ITS/
|Conference Hall 1-2, G/F, Core Building 1, Phase 1, Hong Kong Science Park, Shatin, Hong Kong