Development of a Face Recognition System on an Image Processing LSI Chip

Abstract

We present a real-time and high-precision face recognition system using an image processing LSI chip:Visconti[1]. The system is compact and operates at low power, making it suitable for many purposes, including home security and robot vision. The LSI includes three media processing modules and peripherals which are suitable for machine vision. Face recognition is based on the constrained mutual subspace method (CMSM), implemented on the LSI and optimized to make the best use of the hardware features. The optimization consists of four different levels: instruction, data, task and algorithm. It shows possible implementation of face recognition with multi-core CPUs or other LSI chips. Experimental results show the system operates at 20 frames/sec and a recognition rate is 99.59% when a threshold value is set to 0.55; performance comparable to that of a state-of-the-art system.

Face Recognition with the Multiple Constrained Mutual Subspace Method

“…The eigenvectors, selected in ascending order, are the basis vectors of the constraint subspace. For details of CMSM see [5,7].…”

“…First, an input set of face patterns is obtained from a video sequence. We locate the face pattern from the positions of the pupils and the nostrils obtained automatically by the method described in [1,7]. The pattern is transformed to a vector by raster-scanning of the pattern, and we apply PCA to the vectors to generate an input subspace.…”

“…Through this projection we can extract features that are insensitive to varying facial pose and illumination, while remaining sensitive to change in individual appearance. Using CMSM Sato et al [6] illustrated the effectiveness in a practical security system, while Kozakaya et al [7] demonstrated an implementation of a real-time system on an image processing LSI chip. Although CMSM is effective, a large number of training patterns are required for generating the constraint subspace.…”

Face Recognition from Video by Matching Image Sets

“…Implementations for real-time face recognition from video based on the subspace method were presented in [9,14]. The CMSM method was implemented on a Visconti image processing LSI chip in [14], yielding excellent speed and recognition rates.…”

Face Recognition and Retrieval in Video

“…In CMSM, the test subspace and the reference subspace are projected onto a constraint subspace, where each subspace exhibits small variance and the two subspaces could be better separated. A real-time system implemented using CMSM was demonstrated in [62]. In [82], the authors further introduced the Multiple Constrained Mutual Subspace Method (MCMSM), which generates multiple constraint subspaces by using the ensemble learning algorithms (Bagging and Boosting).…”