•  
  •  
 

ASEAN Journal on Science and Technology for Development

Abstract

In the age of Industry 4.0, recognition of human moving actions becomes an essential element in smart surveillance systems at public places. The requirement of a large number of frames, view dependency and the requirement of big database are still challenging issues for human action recognition in real-time applications. This study proposes an efficient method for basic moving actions, “Walking” and “Running”, based on foot-lift features aiming at the recognition of ongoing moving action by reducing the number of frames required. The plane of moving path and foot-lift are estimated during the action by means of BLOB analysis. The performance of proposed method is evaluated using moving actions from three different datasets. The recognition rate of proposed method is analyzed by using 3, 5 and 7 consecutive frames in each recognition phase. The experimental results show that a minimum accuracy of 80% can be achieved by using only three consecutive frames in both orthogonal and perspective views. When seven frames are used in each recognition phase, the accuracy raises up to 94%. The processing time required for a single frame is 0.08 s or a processing rate of 12 frames per second is achieved. Thus, foot-lift is one reliable feature for human moving action recognition.

Keywords

Human action recognition; Human tracking; Smart surveillance system; Human-computer interaction; Gait features

Received Date

03-dec-2023

Revised Date

05-apr-2024

Accepted Date

28-apr-2024

Download

Included in

Engineering Commons

Share

COinS