Please use this identifier to cite or link to this item:
|Title:||Combined face detection/recognition system for smart rooms|
De Silva, L.C.
|Source:||Kui, J.,De Silva, L.C. (2003). Combined face detection/recognition system for smart rooms. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2688 : 787-795. ScholarBank@NUS Repository.|
|Abstract:||Smart Rooms have many interesting advantages in real world applications. They have cameras, microphones, and other sensors installed for performing different functions such as tracking and recognizing people's expressions and gestures, interpreting their behaviors, and finally extracting the required data for specific purposes. In this paper, we propose an accurate face detection/recognition system to recognize people who enter the smart room, then identifying if he/she is an intruder or a registered user of the facility. Accurate face recognition is still a difficult task, especially in the cases that background, pose, expression, lighting and illumination are unconstrained. Through some experiments, in this paper, we deduce that when taking the central part of the upright frontal faces(including eyes, nose, mouth and chin, but no hair) as samples to make face recognition, the recognition rate will be improved dramatically, even with different expressions, not too extreme lighting change and slight head rotation. For the module of face detection, a support vector machine (SVM) approach is used. And classical eigenface algorithm is utilized to solve the face recognition problem. We combined these two techniques together to construct a system for face detection/recognition with accuracy as high as 96.25%. © Springer-Verlag 2003.|
|Source Title:||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Appears in Collections:||Staff Publications|
Show full item record
Files in This Item:
There are no files associated with this item.
checked on Dec 7, 2017
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.