Deep Learning-Enhanced Motion Recognition for Bionic Prosthetic Arms in Biotechnology Applications

Bochen Zhao,
College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou, China,310000
Haoyu Mao
College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou, China,310000

DOI:https://doi.org/10.5912/jcb1126

Abstract:

This paper introduces a novel deep learning-based approach to enhance synchronized motion recognition in bionic prosthetic arms, crucial for applications in biotechnology. Utilizing an intrinsic rule-based learning framework, the proposed method features hidden units that capture and represent motion patterns between dual bionic arms. The model is designed to predict movements accurately, identifying which input triggers a specific arm's motion and determining the sequence of movements. This approach adapts dynamically to changes in the user's environment, significantly reducing computational complexity compared to traditional kernel density estimation methods. Our experimental validation demonstrates that this deep learning method surpasses existing techniques in accurately detecting and synchronizing movements across varied and challenging environments, including those with high noise levels and significant background interference, such as large shadows. The results reveal that our optimized algorithm improves the precision of movement prediction in prosthetic arms and ensures more natural and responsive user interactions. The enhanced background subtraction method and multi-level background modelling algorithm integrated into our system offer superior performance in managing complex scenarios, making it highly effective for the advanced functionality required in biotechnological applications of bionic prosthetics.