By Topic

An Embedded Robust Facial Feature Detector

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

4 Author(s)
Roux, S. ; France Telecom, Meylan ; Mamalet, F. ; Garcia, C. ; Duffner, S.

In this paper, we present a robust and optimized facial feature detector algorithm which meets the constraints of embedded processors allowing facial feature based services on mobile terminals, such as teleconferencing, advanced user interface, image indexing and security access control. The studied facial feature detector is based on convolutional neural networks, a feature extraction and classification technique which consists of a pipeline of convolution and sub-sampling operations, followed by a multi layer perceptron. The design of embedded systems requires a good trade off between performance and code size due to the limited amount of available resources. We show that such convolutional neural network can efficiently avoid any floating-point computation without any loss in performance. We also propose several algorithmic optimizations to reduce the complexity of the algorithm and avoid common drawbacks of embedded applications such as miss-caches. Experimental results show that our embedded facial feature detection system can accurately locate facial feature with less computational load. It is able to process up to 12 faces/s on an Xscale PXA27x embedded processor @ 624MHz, which represents a speed-up factor of 700 compared to the reference floating point implementation.

Published in:

Machine Learning for Signal Processing, 2007 IEEE Workshop on

Date of Conference:

27-29 Aug. 2007