Papers
6
Total Citations
61
H-Index
6
About
Yusuke Goutsu is a researcher specializing in human motion recognition, gesture analysis, and human-robot interaction, with an emerging focus on egocentric vision and affordance understanding. His foundational work centers on skeleton-based motion recognition, where he pioneered approaches leveraging local body part features, Inverse Kinematics-derived joint relationships, and discriminative body part modeling to classify complex daily human motions. His 2015 contributions introduced multiple kernel learning of Fisher Vectors for skeleton-based systems and a hybrid generative-discriminative gesture recognition framework, establishing robust methodologies that collectively garnered over 24 citations. Alongside this, his multi-modal gesture recognition work demonstrated the value of integrating motion, audio, and video streams for improved accuracy. His 2017 research extended motion classification to multi-class scenarios with natural language sentence descriptions, bridging motion analysis and language generation — a direction he had explored as early as 2013 through large-scale N-gram-based motion-to-sentence systems. More recently, his 2023 work on fine-grained affordance annotation for egocentric hand-object interaction videos, already accumulating 12 citations, signals a significant pivot toward action anticipation and robot imitation learning, highlighting his continued relevance at the intersection of computer vision and intelligent robotics.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6