It has been successfully applied to modeling the visual receptive fields of the cortical neurons. Slow feature analysis sfa extracts slowly varying features from a quickly varying input signal. Slow feature analysis sfa was first proposed in 25. Probabilistic slow features for behavior analysis sewa project. Recently, we have seen huge advances in action detection or recognition in well controlled e. Slow feature analysis for human action recognition abstract. Action recognition using spatialoptical data organization and sequential learning framework yuan yuana, yang zhaoa,b, qi wangc, acenter for optical imagery analysis and learning optimal, xian institute of optics and precision mechanics of cas, xian 710119, china buniversity of chinese academy of sciences, beijing 49, china cschool of computer science, center for optical imagery. Slow feature analysis sfa extracts slowly varying features from a quickly varying input signal 1. Slow feature analysis for human action recognition ieee xplore. Human action recognition using modified slow feature.
Another important aspect of the status on human action recognition research is that most studies concentrated on human action feature representations. Recently, we have seen huge advances in action detection or recognition in wellcontrolled e. We propose to generalize slow feature analysis to steady feature analysis. Unlike slow feature analysis, we redefine the objective function with supervised information, which make the modified sfa more suitable to preserve the slow feature and label. Action recognition skeleton joint stream multiorder streams slow feature analysis. Finally, an incremental sfa algorithm for change detection. Perceptual principles for video classification with slow feature.
Firstly, we use a modified slow feature analysis sfa to extract video local feature. Abstractslow feature analysis sfa extracts slowly varying features from a quickly varying input signal 1. Slow feature analysis for human action recognition deepai. Sufficient experimental results in neuroscience suggest that the temporal slowness principle is a general learning principle in visual perception. Learning skeleton stream patterns with slow feature analysis for. Even with the manual assistance in selecting features and methods, accurate human action recognition is still a highly cumbersome task due to complex. Slow feature analysis for human action recognition. A novel human action recognition method is proposed, which includes two periods of action feature extraction and action recognition.
A dual fast and slow feature interaction in biologically inspired visual recognition of human action article pdf available in applied soft computing 62 september 2015 with 245 reads. Slow feature analysis for human action recognition arxiv. Were upgrading the acm dl, and would like your input. To solve this problem, in this work, we summarize human action recognition methods applicable to different types of data and involving handcrafted feature based and feature learning methods. Human action detection by boosting efficient motion features. Slow feature analysis for human action recognition ieee. Human action recognition using slow feature analysis.
1111 729 351 962 104 706 314 1321 168 205 372 721 499 248 91 1174 1034 891 1347 472 1071 559 501 1082 245 795 372 926 1391 513 312 812 573 1331