Yu Zhu

Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Guodong Guo

Committee Co-Chair

Donald A Adjeroh

Committee Member

Hany H Ammar

Committee Member

Hong-Jian Lai

Committee Member

Xin Li


Human action recognition is very useful in many applications in various areas, e.g. video surveillance, HCI (Human computer interaction), video retrieval, gaming and security. Recently, human action recognition becomes an active research topic in computer vision and pattern recognition. A number of action recognition approaches have been proposed. However, most of the approaches are designed on the RGB images sequences, where the action data was collected by RGB/intensity camera. Thus the recognition performance is usually related to various occlusion, background, and lighting conditions of the image sequences. If more information can be provided along with the image sequences, more data sources other than the RGB video can be utilized, human actions could be better represented and recognized by the designed computer vision system.;In this dissertation, the multi-modality human action recognition is studied. On one hand, we introduce the study of multi-spectral action recognition, which involves the information from different spectrum beyond visible, e.g. infrared and near infrared. Action recognition in individual spectra is explored and new methods are proposed. Then the cross-spectral action recognition is also investigated and novel approaches are proposed in our work. On the other hand, since the depth imaging technology has made a significant progress recently, where depth information can be captured simultaneously with the RGB videos. The depth-based human action recognition is also investigated. I first propose a method combining different type of depth data to recognize human actions. Then a thorough evaluation is conducted on spatiotemporal interest point (STIP) based features for depth-based action recognition. Finally, I advocate the study of fusing different features for depth-based action analysis. Moreover, human depression recognition is studied by combining facial appearance model as well as facial dynamic model.