PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 16856670

Face recognition using kernel scatter-difference-based discriminant analysis.

Qingshan Liu·Xiaoou Tang·Hanqing Lu·Songde Ma
View original source ↗Export metadata ↓

Abstract

There are two fundamental problems with the Fisher linear discriminant analysis for face recognition. One is the singularity problem of the within-class scatter matrix due to small training sample size. The other is that it cannot efficiently describe complex nonlinear variations of face images because of its linear property. In this letter, a kernel scatter-difference-based discriminant analysis is proposed to overcome these two problems. We first use the nonlinear kernel trick to map the input data into an implicit feature space F. Then a scatter-difference-based discriminant rule is defined to analyze the data in F. The proposed method can not only produce nonlinear discriminant features but also avoid the singularity problem of the within-class scatter matrix. Extensive experiments show encouraging recognition performance of the new algorithm.

Explore related subjects

Discriminant AnalysisFaceLinear ModelsPattern Recognition, AutomatedLetterResearch Support, Non-U.S. Gov't

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Qingshan Liu, Xiaoou Tang, Hanqing Lu, Songde Ma. 2006. Face recognition using kernel scatter-difference-based discriminant analysis.. https://doi.org/10.1109/tnn.2006.875970

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2007

Determination of the geographic origin of onions between three main production areas in Japan and other countries by mineral composition.

Onions (Allium cepa L.) are produced in many countries and are one of the most popular vegetables in the world, thus leading to an enormous amount of international trade. It is currently important that a scientific technique be developed for determining geographic origin as a means to detect fraudulent labeling. We have therefore developed a technique based on mineral analysis and linear discriminant analysis (LDA). The onion samples used in this study were from Hokkaido, Hyogo, and Saga, which are the primary onion-growing areas in Japan, and those from countries that export onions to Japan (China, the United States, New Zealand, Thailand, Australia, and Chile). Of 309 samples, 108 were from Hokkaido, 52 were from Saga, 77 were from Hyogo, and 72 were from abroad. Fourteen elements (Na, Mg, P, Mn, Co, Ni, Cu, Zn, Rb, Sr, Mo, Cd, Cs, and Ba) in the samples were determined by frame atomic adsorption spectrometry, inductively coupled plasma optical emission spectrometry, and inductively coupled plasma mass spectrometry. The models established by LDA were used to discriminate the geographic origin between Hokkaido and abroad, Hyogo and abroad, and Saga and abroad. Ten-fold cross-validations were conducted using these models. The discrimination accuracies obtained by cross-validation between Hokkaido and abroad were 100 and 86%, respectively. Those between Hyogo and abroad were 100 and 90%, respectively. Those between Saga and abroad were 98 and 90%, respectively. In addition, it was demonstrated that the fingerprint of an element pattern from a specific production area, which a crop receives, did not easily change by the variations of fertilization, crop year, variety, soil type, and production year if appropriate elements were chosen.

Discriminant Analysis↗
PubMed2006

Application of multivariate statistical methods to water quality assessment of the watercourses in Northwestern New Territories, Hong Kong.

Multivariate statistical methods, i.e., cluster analysis (CA) and discriminant analysis (DA), were used to assess temporal and spatial variations in the water quality of the watercourses in the Northwestern New Territories, Hong Kong, over a period of five years (2000-2004) using 23 parameters at 23 different sites (31,740 observations). Hierarchical CA grouped the 12 months into two periods (the first and second periods) and classified the 23 monitoring sites into three groups (group A, group B, and group C) based on similarities of water quality characteristics. DA provided better results with great discriminatory ability for both temporal and spatial analysis. DA also provided an important data reduction because it only used six parameters (pH, temperature, five-day biochemical oxygen demand, fecal coliforms, Fe, and Ni) for temporal analysis, affording about 84% correct assignations, and seven parameters (pH, ammonia-nitrogen, nitrate nitrogen, fecal coliforms, Fe, Ni, and Zn) for spatial analysis, affording more than 90% correct assignations. Therefore, DA allowed a reduction in the dimensionality of the large data set and indicated a few significant parameters that were responsible for most of the variations in water quality. Thus, this study demonstrated that the multivariate statistical methods are useful for interpreting complex data sets in the analysis of temporal and spatial variations in water quality and the optimization of regional water quality monitoring network.

Discriminant Analysis↗
PubMed2006

Changes in volatile and phenolic compounds with malaxation time and temperature during virgin olive oil production.

Virgin olive oils produced at wide ranges of malaxation temperatures (15, 30, 45, and 60 degrees C) and times (30, 60, 90, and 120 min) in a complete factorial experimental design were discriminated with stepwise linear discriminant analysis (SLDA) revealing differences with processing conditions. Virgin olive oils produced at 15 and 60 degrees C for 30 min showed the most significant (p < 0.01) differences. Discrimination was based upon volatile and phenolic compounds detected in olive oils, peroxide value (PV), free fatty acids (FFA), ultraviolet (UV) absorbances, and oil yield. There were different discriminating variables for processing conditions illustrating the dependence of virgin olive oil quality on malaxation time and temperature. Volatile compounds were the dominant discriminating variables. Common oxidation indicators of olive oil (PV, K232, and K270) were not among the variables that significantly (p < 0.01) changed with malaxation time and temperature. Variables that discriminated both malaxation time and temperature were hexanal, 3,4-dihydroxyphenyl ethyl alcohol-decarboxymethyl elenolic acid dialdehyde (3,4-DHPEA-DEDA) and FFA, whereas 1-penten-3-ol, E-2-hexenal, octane, tyrosol, and vanillic acid significantly (p < 0.01) changed with temperature only and Z-2-penten-1-ol, (+)-acetoxypinoresinol, and oil yield changed with time only. Virgin olive oil quality was significantly influenced by malaxation temperature, whereas oil yield discriminated malaxation time. This study demonstrates the two modes of hexanal formation: enzymatic and nonenzymatic during virgin olive oil extraction.

Discriminant Analysis↗