ITS Campus, ITS News – Research on Human Motion Analysis (HMA) is now advancing in several fields, including healthcare. A new doctoral graduate from Institut Teknologi Sepuluh Nopember (ITS), Dr. Ir. Endang Sri Rahayu, M.Kom, has developed an innovative method to recognize human activities by observing joint points using Deep Convolutional Neural Networks (DCNN), a method within Artificial Intelligence (AI).
During her open doctoral defense held at the ITS’ Department of Electrical Engineering on Monday (12/8), Endang emphasized the importance of research in human activity recognition. This research supports medical rehabilitation processes, elderly activity monitoring, and the development of intelligent robot movements. “To address various human activities, the research needs to be developed using various methods to achieve high accuracy,” she explained.
Endang further highlighted that recognizing human activity is critical in healthcare because it can detect abnormal movements. Such abnormal movements could serve as an indicator to monitor disease risks, especially in the elderly. “Therefore, research on human activity observation becomes essential,” said the lecturer from Jayabaya University’s Department of Electrical Engineering, Jakarta.
Driven by this problem, Endang’s dissertation titled A Combination Model of Joint Angle Shifts with Deep Learning to Recognize Human Activities aims to recognize human activities by extracting joint features. “This research analyzes joint positions using the DCNN model,” explained the Jombang-born researcher, who was born on April 27, 1965.
Using the Florence 3D Actions dataset, this research observed 15 human joint points, including the head, shoulders, and ankles, which serve as indicators of human movement. Joints were chosen as movement indicators because they connect the human skeleton and move following human activity patterns. “Thus, joints are the ideal indicators as joint points represent human activity patterns,” explained the ITS Electrical Engineering graduate.
Endang added that the distance between joints is calculated frame by frame from a segment of observation video using the Euclidean distance technique. However, this method struggled to distinguish some human activities, such as sitting and standing. “This occurs because the joint distance-based method only considers absolute joint position changes without accounting for movement direction,” she explained.
Delving deeper, Endang noted the need for a method that could differentiate between activities with similar movement distances. She introduced a joint angle shift method that successfully distinguished activities like sitting and standing, which have similar joint position changes but differ in movement direction.
Reporter: Mohammad Febryan Khamim
ITS Campus, ITS News —Losing a pet is a sad thing for its owners. To anticipate this, a team
ITS Campus, ITS News — Institut Teknologi Sepuluh Nopember (ITS) continues to prove itself as a home for talented
ITS Campus, ITS News — Along with the development of technology, the construction sector has also experienced rapid growth
ITS Campus, ITS News — Proving itself as a home for champions, Institut Teknologi Sepuluh Nopember (ITS) managed to