Human Action Recognition Based on Embedded HMM
Tanvir Fatima Naik Bukht, Ahmad Jalal
Abstract
One of the most active subfields of computer vision and pattern recognition, human action recognition seeks to identify specific actions in still photos. The wide range of possible uses in this area has attracted a lot of interest from researchers in fields like human-computer interaction, video analysis, and surveillance. Researchers have developed a variety of algorithms and methods to accurately identify and classify the various actions that individuals performed while they were captured in photographs. An embedded HMM is going to be utilised in the process of developing an action recognition technique that will be presented in this article. It is recommended that the HSV colour transformation be carried out first in order to enhance the clarity of the video frames. The MOT and Vibe methods are used to get the silhouette. Feature extraction is accomplished through the utilisation of Texton Map and ORB. The fuzzy optimisation process is then utilised for the purpose of feature discrimination. After that, the features are input into an embedded HMM, where they are categorised into specific human actions based on the final qualities that they possess. The UT interaction data that was used in the experiment was recognised 93.8%.