Face Identification and Verification in Wearable devices
<< Back to Research Areas & Projects
Our focus is on fundamental representations of face images and feature extraction that do largely statistical pattern analysis and machine learning. We have proposed a subclass based whole space discriminant analysis method that helps in finding stable low dimensional feature representation of face images. These features are used to identify and/or verify one person from others.
Fig. 1. Algorithm Framework
In this work, we propose to divide each class (a person) into subclasses using spatial partition trees (as shown in Fig. 1) which helps in better capturing the intra-personal variances arising from the appearances of the same individual. We perform a comprehensive analysis on within-class and within-subclass eigenspectrums of face images and propose a novel method of eigenspectrum modeling which extracts discriminative features of faces from both within-subclass and total or between-subclass scatter matrices. Effective low-dimensional face discriminative features are extracted for face recognition (FR) after performing discriminant evaluation in the entire eigenspace. Experimental results on popular face databases and the challenging unconstrained YouTube Face database show the superiority of our proposed approach on all three databases.
Apart from desktop and laptop applications, we have implemented this method in mobile and wearable devices like Google Glass (Fig. 2(a)) and Narrative Clip 2 (Figs. 2(b)-(c)). It runs as a real-time standalone system in Google Glass and also client-server mode, connecting a phone via Bluetooth. A sample image captured by Google Glass is shown below in Fig. 3.
Fig. 3. Faces are being detected and recognized using Google Glass. Red box shows the person identification along with the names. (click on the image for zooming in).
Red boxes show the recognized faces (people registered in the database), blue box shows the unknowns (people who are not in the database).
Watch a small video clip here [AVI].
If you want to obtain a copy of the wearable device database please contact us at: bmandal AT i2r dot a-star dot edu dot sg.
Fig. 4 shows a snapshot using the webcam of a notebook. Any person can be registered (one-time) into the system and then he/she can be identified and/or verified from at a distance (totally non-intrusive).
Fig. 4. Face recognition using a webcam and notebook.
Red box shows the person identification along with the names,
blue box shows Unknown. (click on the image for zooming in).
Read details here:
In the News [2015]: Gov site - Smart glasses for seniors with dementia
(http://www.gov.sg/news/content/the-straits-times-smart-glasses-for-seniors-with-dementia) [Picture]
(The Strait Times [2015]: http://www.straitstimes.com/singapore/health/smart-glasses-for-seniors-with-dementia)
(MIT CBMM [2015]: http://cbmm.mit.edu/news-events/news/smart-glasses-seniors-dementia-straits-times)
(BBC News [2015]: http://www.bbc.com/news/technology-33704192)
(AsiaOne [2015]: http://news.asiaone.com/news/singapore/smart-glasses-seniors-dementia)
(Mypaper [2015]: http://mypaper.sg/top-stories/smart-glasses-guide-seniors-works-20150710)
[J2] Q. Xu, S. Ching, B. Mandal, L. Li, J-H Lim, M. Mukawa and C. Tan, “SocioGlass: Social interaction assistance with face recognition on google glass,” Journal of Scientific Phone Apps and Mobile Devices, Springer, Mar 2016. (Accepted) [PDF]
[J1] B. Mandal, Wang Zhikai, Liyuan Li and A. Kassim, “Performance evaluation of local descriptors and distance measures on benchmarks and first-person-view videos for face identification,” Journal of Neurocomputing, Jul 2015. (Accepted) (Impact factor: 2.292) [PDF]
[C4] B. Mandal, Liyuan Li, V. Chandrasekha and Joo Hwee Lim, "Whole Space Subclass Discriminant Analysis for Face Recognition", IEEE International Conference on Image Processing (ICIP), pp. 329-333, Quebec city, Canada, Sep 2015. [PDF]
[C3] S. Ching, B. Mandal, Q. Xu, Li Liyuan and J-H. Lim, "Enhancing Social Interaction with Seamless Face Recognition on Google Glass: Leveraging opportunistic multi-tasking on smart phones", 17th International Conference on Human-Computer Interaction with Mobile Devices and Services (MobileCHI) Aug, 2015 Copenhagen, Denmark. (Poster) [PDF]
[C2] B. Mandal, S. Ching, Li Liyuan, V. Chandrasekha, C.Tan and J-H. Lim, "A Wearable Face Recognition System on Google Glass for Assisting Social Interactions", 3rd International Workshop on Intelligent Mobile and Egocentric Vision, ACCV, Singapore, pp. 419-433, 1-5 Nov 2014. (Oral) [PDF]
[C1] B. Mandal, Wang Zhikai, Liyuan Li and Ashraf A. Kassim, "Evaluation of Descriptors and Distance Measures on Benchmarks and First-Person-View Videos for Face Identification", International Workshop on Robust Local Descriptors for Computer Vision, ACCV, Singapore, pp. 585-599, 1-5 Nov 2014. (Oral) [PDF]