SubXPCA and a generalized feature partitioning approach to principal component analysis

Kadappa, Vijayakumar; Negi, Atul

doi:10.1016/j.patcog.2007.08.006

Cited by 39 publications

(6 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To exploit both local and global information, SubXPCA [11] and FLPCA [12] were proposed; they were proven to be better as compared to PCA, 2DPCA and other methods. In this paper, we exploit the benefits of both DWT and SubXPCA by proposing a hybrid method.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An improved feature extraction method based on DWT and 2DSubXPCA methods

Devi¹,

Rao²

2015

2015 International Conference on Computing and Network Communications (CoCoNet)

View full text Add to dashboard Cite

Face is one of the popular biometric used in human authentication for providing secured access. Feature reduction is a challenging task for fast and efficient recognition. Many techniques like Discrete wavelet transform (DWT), Principal component analysis (PCA) and Linear discriminant analysis (LDA) have been used for feature reduction. Some issues like pose variation, change in facial expression and poor illumination make the recognition process a difficult task. In this paper, we bring in a hybrid approach based on DWT and 2DSubXPCA for feature extraction and use nearest neighbor for classification. A maximum classification accuracy of 97.5% with highest dimensionality reduction is obtained for ORL database.

show abstract

Section: Introductionmentioning

confidence: 99%

“…Section 4 discusses the experimental results followed by conclusion in section 5. In this section, we outline DWT [2], 2DPCA [3], 2D version of original SubPCA [5] and 2D Cross SubPCA [11] methods.…”

Section: Introductionmentioning

confidence: 99%

An improved feature extraction method based on DWT and 2DSubXPCA methods

Devi¹,

Rao²

2015

2015 International Conference on Computing and Network Communications (CoCoNet)

View full text Add to dashboard Cite

show abstract

“…Many approaches such as neural network based PCA methods [6], 2DPCA based methods [9] reduce computational complexity as compared to classical PCA method. However, these methods are based on whole-patterns, which are suitable for global feature extraction like classical PCA, and they may not perform well if local variations are prominent [8]. It is known that PCA performs global feature extraction that retains information based upon covariances between every pair of original features.…”

Section: Introductionmentioning

confidence: 99%

“…Recently, we proposed a novel FP-PCA method, SubXPCA [8], which extracts both global and local feature variations. SubXPCA is computationally more efficient as compared to PCA, and also effective in classification accuracy.…”

Section: Introductionmentioning

confidence: 99%

SubXPCA versus PCA: A Theoretical Investigation

Negi

Kadappa²

2010

2010 20th International Conference on Pattern Recognition

Self Cite

View full text Add to dashboard Cite

Principal Component Analysis (PCA) is a widely accepted dimensionality reduction technique that is optimal in a MSE sense. PCA extracts 'global' variations and is insensitive to 'local' variations in subpatterns.Recently, we have proposed a novel approach, SubX-PCA, which was more effective computationally than PCA and also effective in computing principal components with both global and local information across subpatterns. In this paper, we show the near-optimality of SubXPCA (in terms of summarization of variance) by proving analytically that 'SubXPCA approaches PCA with increase in number of local principal components of subpatterns.' This is demonstrated empirically upon CMU Face Data.

show abstract

“…Sub-XPCA was applied on non-image data and the technique showed that it is not always correct to rely only on locally-extracted features from feature blocks (as done by methods like modPCA) but a sophisticated combination approach with global features using inter-block feature correlations is required. A more comprehensive study of the feature partitioning approach is presented in (Negi and Vijaya Kumar, 2007).…”

Section: Introductionmentioning

confidence: 99%