Visual object recognition using probabilistic kernel subspace similarity

Lee, Jianguo; Wang, Jingdong; Zhang, Changshui; Bian, Zhaoqi

doi:10.1016/j.patcog.2005.01.007

Cited by 7 publications

(3 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Further studies lead to a modified decomposition called smooth orthogonal decomposition (SOD) [28][29][30], and it can be regarded as a projection of an ensemble of spatially distributed data. In smooth local decomposition, the one dimensional time series is reconstructed to high dimensional phase space, then noises in the data can be reduced by being projected to the tangent subspace [31]. The vector directions of the projection can make sure the variance is as maximum as possible, and the motions obtaining along these vector directions are the smoothest in terms of time.…”

Section: Introductionmentioning

confidence: 99%

Fault Classification of Rotary Machinery Based on Smooth Local Subspace Projection Method and Permutation Entropy

Xiao

2019

Applied Sciences

View full text Add to dashboard Cite

Collected mechanical signals usually contain a number of noises, resulting in erroneous judgments of mechanical condition diagnosis. The mechanical signals, which are nonlinear or chaotic time series, have a high computational complexity and intrinsic broadband characteristic. This paper proposes a method of gear and bearing fault classification, based on the local subspace projection noise reduction and PE. A novel nonlinear projection noise reduction method, smooth orthogonal decomposition (SOD), is proposed to denoise the vibration signals of various operation conditions. SOD can decompose the reconstructed multiple strands to identify smooth local subspace. In the process of projection from a high dimension to a low dimension, a new weight matrix is put forward to achieve a better denoising effect. Afterwards, permutation entropy (PE) is applied in the detection of time sequence randomness and dynamic mutation behavior, which can effectively detect and amplify the variation of vibration signals. Hence PE can characterize the working conditions of gear and bearing under different conditions. The experimental results illustrate the effectiveness and superiority of the proposed approach. The theoretical derivations, numerical simulations and experimental studies, all confirm that the proposed approach based on the smooth local subspace projection method and PE, is promising in the field of the fault classification of rotary machinery.

show abstract

Section: Introductionmentioning

confidence: 99%

Fault Classification of Rotary Machinery Based on Smooth Local Subspace Projection Method and Permutation Entropy

Xiao

2019

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…In [6], Wang et al proposed a unified subspace analysis technique for face recognition. Lee et al [7] proposed kernel extensions to the similarity measure used in [8]. In [9], Ramanathan et.al used the face similarity measure proposed in [8] for face verification across age.…”

Section: Introductionmentioning

confidence: 99%

Generalized subspace based high dimensional density estimation

Vadivel

Sargin

Joshi

et al. 2011

2011 18th IEEE International Conference on Image Processing

View full text Add to dashboard Cite

Our paper presents a novel high dimensional probability density estimation technique using any dimensionality reduction method. Our method first performs subspace reduction using any matrix factorization algorithm and estimates the density in the low-dimensional space using sample-point variable bandwidth kernel density estimation. Subsequently, the high dimensional density is approximated from the low dimensional density parameters. The reconstruction error due to dimensionality reduction process is also modeled in a principled and efficient manner to obtain the high dimensional density estimate. We show the effectiveness of our technique by using two popular dimensionality reduction tools, principal component analysis and non-negative matrix factorization. This technique is applied to AT&T, Yale, Pointing'04 and CMU-PIE face recognition datasets and improved performance compared to other dimensionality reduction and density estimation algorithms is obtained.

show abstract

“…As a result, it will be shown that, with the proposed face representation, better classification results than the traditional Fisherfaces and than Kittler's and Fagertun's representations are obtained. It will also be shown that, similar to other algorithms that aim at characterizing intra-personal and extra-personal face differences (Shen and Bai, 2006;Lee et al, 2005), the algorithm used to obtain the proposed face representation allows us to visualize what the individual's most discriminative features are.…”

Section: Introductionmentioning

confidence: 99%