Spatio-temporal action localization and detection for human action recognition in big dataset

3D Features for human action recognition with semi‐supervised learning

“…The proposed method is compared with the previous methods [1–5, 30] for KTH dataset. The confusion matrix for the dataset is given in Table 4, and the comparison with other state‐of‐the‐art methods is shown in Table 5.…”

“…These 2D features are less effective to differentiate action classes which are similar and differentiated by speed of action. Haar wavelet transform is used as feature in [3] and trajectory‐based approach is used in [30]. All these techniques do not take any special care for closely related actions like ‘jog’ and ‘run’.…”

A Fine Grainedresearch Over Human Action Recognition

“…Among various available traditional classifiers, Support Vector Machine [54,56,58,59,60,62,63]and random forest [65,66] has gained an excellent performance in the classification of different actions for all types of action videos. K. G. C. Manocha, R. Rodrigo [56] accomplished the SVM algorithm for action recognition.…”

“…Experiments are conducted on the standard datasets such as KTH and UCF101 and HMDB51 datasets. H. Zhang et al [65] developed an action recognition method which combines the sparse coding with gradient information. In this approach, initially, the depth of gradient information and distance between the joints of the 3D skeleton are extracted to find the coarse depth-skeleton features.…”

Human Action Recognition Based on RGB-D and Local Interactive Regions Detection

“…Human action recognition is an active research topic in computer vision for its wild application in human-computer interaction and intelligent surveillance [1][2][3]. In the past few decades, many methods have been developed to recognize different actions from video sequences captured by visible light cameras and have been applied into some practical application.…”