A Visual-Based Approach for Manual Operation Evaluation

Yiyao Zhao,Zhen Wang,Yanyu Lu,Shan Fu
DOI: https://doi.org/10.1007/978-3-030-49044-7_23
2020-01-01
Abstract:In order to improve the human-machine interface design and monitor the operational performance, hand behavior detection and analysis are very important. However, there is a lack of effective indicators of hand operation behavior. This paper introduces a vision-based approach for manual operation evaluation. K-curvature is employed to detect the operation key points on hand and the palm center is located combining image region moments and palm maximum circle fitting algorithm. The inter-frame velocity cosine is used as a representation of the motion sequence and we extracted the features of approximate entropy (ApEn) and maximum duration ratio (MDR) from the sequences to distinguish different operation statuses. A digit clicking experiment with different difficulty is designed to study operation behavior characteristics and we found the features of ApEn and MDR had a significant difference between position-determined operation and position-unknown operation. When the operation process is aimless and inconsistent, the ApEn value tends to be larger and the MDR value tends to be smaller. Our approach could effectively describe hand operation and assist manual operation performance evaluation and usability improvement.
What problem does this paper attempt to address?