Institution profile

University of Tsukuba

Academic institutionasia · jp
Official website
Research library220linked papers
Opportunities0open roles
Selected work

Representative Papers

Laugh at Your Own Pace: Basic Performance Evaluation of Language Learning Assistance by Adjustment of Video Playback Speeds Based on Laughter Detection

Jun 01, 2022ACM Conference on Learning @ Scale

Second-language learners often struggle to comprehend native-speech videos played at natural speed, limiting the efficacy of extensive viewing. To address this, we propose a non-intrusive, adaptive playback rate control method grounded in spontaneous laughter detection. Our approach leverages real-time audio spectrogram analysis and temporal modeling to identify viewers’ natural laughter responses to comedic content—serving as implicit, zero-effort comprehension feedback—and dynamically adjusts playback speed without requiring manual interaction, speech input, or textual annotation. The system integrates a laughter-driven adaptive controller with a TOEIC-based proficiency stratification framework. Empirical evaluation demonstrates that the method significantly improves comprehension accuracy for learners scoring below 700 on the TOEIC, thereby expanding their access to authentic native-speed video resources. This work represents the first application of spontaneous laughter as a real-time, implicit signal for adaptive language comprehension support, establishing a novel paradigm for unobtrusive, marker-free, personalized audiovisual learning.

6 citationsRead paper

AR Object Layout Method Using Miniature Room Generated from Depth Data

Jan 07, 2026ICAT-EGVE

This work addresses the lack of intuitive and efficient interaction techniques for positioning and scaling virtual objects in augmented reality (AR), which hinders layout efficiency and user experience. It introduces, for the first time, the World-in-Miniature (WIM) paradigm—commonly used in virtual reality—into AR environments. By leveraging the device’s depth sensor to reconstruct a real-time 3D mesh of the surrounding room, the system dynamically generates an interactive miniature model of the physical space. Users can directly manipulate virtual objects within this scaled-down representation to control their real-world counterparts with high spatial precision. Although the approach does not reduce overall task completion time, it significantly alleviates physical and cognitive workload, markedly enhancing comfort, controllability, and interaction intuitiveness during the layout process.

1 citationsRead paper

MGP-KAD: Multimodal Geometric Priors and Kolmogorov-Arnold Decoder for Single-View 3d Reconstruction in Complex Scenes

Sep 14, 2025International Conference on Information Photonics

Single-view 3D reconstruction in complex real-world scenes is often hindered by noise, object diversity, and data scarcity, leading to insufficient geometric accuracy and detail recovery. To address these challenges, this work proposes the MGP-KAD framework, which first generates category-level geometric priors through clustering and dynamically fuses RGB images with multimodal geometric cues. Furthermore, it introduces a hybrid decoder based on Kolmogorov–Arnold Networks (KANs), overcoming the representational limitations of conventional linear decoders when handling complex multimodal inputs. This approach marks the first application of KANs to 3D reconstruction and achieves state-of-the-art performance on Pix3D, significantly enhancing geometric completeness, surface smoothness, and fine-detail preservation.

1 citationsRead paper
Recent publications

Latest Papers