LF-EME: Local features with elastic manifold embedding for human action recognition

Deng, Xiaoyu; Liu, Xiao; Song, Mingli; Cheng, Jun; Bu, Jiajun; Chen, Chun

doi:10.1016/j.neucom.2012.06.011

Cited by 17 publications

(12 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The experiments on the KTH and UCF Sports datasets show that our method achieves better performance than the classical methods published recently [11][12][13][14][15][16].…”

Section: Introductionmentioning

confidence: 80%

“…Table 1 shows the performance comparison between our system and some classical system published recently. The competing methods include local representation-based approaches [11][12][13][14][15][16], global representation-based approach. In detail, SC was used for feature coding together with BoF in [11], local feature distribution information was used in [14], ST context feature was employed in [12], sparse representation-based classification methods was applied in [13], the global representation method was adopted in [15], and a framework based on Elastic Manifold Embedding together with local interest point features to handle human action recognition in [16].…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Action Recognition Using Hierarchical STIP Saliency and Mixed Neighborhood Features

Yang¹,

Ma²

2016

IJCA

View full text Add to dashboard Cite

show abstract

“…The experiments on the KTH and UCF Sports datasets show that our method achieves better performance than the classical methods published recently [11][12][13][14][15][16].…”

Section: Introductionmentioning

confidence: 80%

Section: Resultsmentioning

confidence: 99%

Action Recognition Using Hierarchical STIP Saliency and Mixed Neighborhood Features

Yang¹,

Ma²

2016

IJCA

View full text Add to dashboard Cite

show abstract

“…And the motion information (described as STOE feature) shows how the body part moves. It is noted that our feature do not contain appearance information (for example, histogram of oriented gradient (HOG) feature), because appearance information [18] 2010 Split 94.9 84.3 ---Wu et al [35] 2011 LOOCV 94.5 91.3 ---Escobar et al [15] 2012 Split 90.6 ---99.2 Guha et al [19] 2012 LOOCV ---91.1 98.9 Bregonzio et al [36] 2012 LOOCV 94.3 ---96.6 Zhang et al [38] 2012 LOOCV 95.6 87.3 ---Saghafi et al [16] 2012 LOOCV 92.6 ---100 Deng et al [9] 2012 LOOCV 96.9 88.4 100 LGSR+MSPC-LC LOOCV 98.5 93.5 100 Table 6. Confusion matrix on the UCF Sports dataset with 4-level spatial pyramid match (SPM).…”

Section: Lgsr Versus Srcmentioning

confidence: 99%

“…The quality of the learnt model decides discriminative power usually results in good classification result. There are two representation methods: holistic action representations [1][2][3][4][5][6][7] and local action representations [8][9][10][11][12][13][14][15][16][17][18].…”

Section: Introductionmentioning

confidence: 99%

Multiscale Spatial Position Coding under Locality Constraint for Action Recognition

Yang¹,

Ma²,

Xie³

2015

Journal of Electrical Engineering and Technology

View full text Add to dashboard Cite

-In the paper, to handle the problem of traditional bag-of-features model ignoring the spatial relationship of local features in human action recognition, we proposed a Multiscale Spatial Position Coding under Locality Constraint method. Specifically, to describe this spatial relationship, we proposed a mixed feature combining motion feature and multi-spatial-scale configuration. To utilize temporal information between features, sub spatial-temporal-volumes are built. Next, the pooled features of sub-STVs are obtained via max-pooling method. In classification stage, the LocalityConstrained Group Sparse Representation is adopted to utilize the intrinsic group information of the sub-STV features. The experimental results on the KTH, Weizmann, and UCF sports datasets show that our action recognition system outperforms the classical local ST feature-based recognition systems published recently.

show abstract

“…However, the local features usually reside on nonlinear manifolds [20]. Neither SVQ nor SC can preserve the nonlinear manifold structure.…”

Section: Introductionmentioning

confidence: 99%

Action Recognition Using Weighted Locality-Constrained Linear Coding

Yang

2015

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYRecently, locality-constrained linear coding (LLC) as a coding strategy has attracted much attention, due to its better reconstruction than sparse coding and vector quantization. However, LLC ignores the weight information of codewords during the coding stage, and assumes that every selected base has same credibility, even if their weights are different. To further improve the discriminative power of LLC code, we propose a weighted LLC algorithm that considers the codeword weight information. Experiments on the KTH and UCF datasets show that the recognition system based on WLLC achieves better performance than that based on the classical LLC and VQ, and outperforms the recent classical systems.

show abstract

LF-EME: Local features with elastic manifold embedding for human action recognition

Cited by 17 publications

References 60 publications

Action Recognition Using Hierarchical STIP Saliency and Mixed Neighborhood Features

Action Recognition Using Hierarchical STIP Saliency and Mixed Neighborhood Features

Multiscale Spatial Position Coding under Locality Constraint for Action Recognition

Action Recognition Using Weighted Locality-Constrained Linear Coding

Contact Info

Product

Resources

About