Sub-Sampled Dictionaries for Coarse-to-Fine Sparse Representation-based Human Action Recognition. Spotlight presentation for the main track of ICME 2014.
Sub-sampled dictionaries for coarse-to-fine sparse representation-based human action recognition
1. Poster Spotlights
SUB-SAMPLED DICTIONARIES FOR COARSE-TO-FINE SPARSE
REPRESENTATION-BASED HUMAN ACTION RECOGNITION
Poster Session 5, July 17h
JongHo Lee, Hyun-seok Min, Jeong-jik Seo,
Wesley De Neve, and Yong Man Ro
506
2. SUB-SAMPLED DICTIONARIES FOR COARSE-TO-FINE
SPARSE REPRESENTATION-BASED HUMAN ACTION RECOGNITION
1. Introduction
Sparse representation-based classification
(SRC) has recently attracted much attention
However, the computational complexity of SRC
makes its usage challenging in practice
We propose a novel method for human action
recognition, leveraging coarse-to-fine sparse
representations
2. Proposed Method
The time complexity of SRC depends on the
dictionary size
4. Conclusions
We proposed a novel method for human
action recognition using coarse-to-fine sparse
representations
This proposed method is able to achieve
efficient human action recognition with no
substantial loss in accuracy
3. Experimental Results
A: Conventional SRC (using only the Fine-Grained Dictionary)
B: Proposed Method (using Coarse-to-Fine Representations)
1 2 3 4 K…
1 2 3 4 K…
1 4 H…
Coarse-Grained Dictionary
Fine-Grained Dictionary
Pruned Fine-Grained Dictionary
Candidate
Classes
67.5
32.8
0.0
20.0
40.0
60.0
80.0
A B
Timecomplexity
Time Complexity (s)
0.8438
0.8567
0.835
0.84
0.845
0.85
0.855
0.86
A B
RecognitionAccuracy
Accuracy