TY  - JOUR
AU  - Shang, Wenli
AU  - Alena, Kazlova
PY  - 2026
DA  - 2026/09/26
TI  - Adaptive Field-of-View Restriction Based on Sigmoid and Bézier Smoothing Control Functions
JO  - Chinese Journal of Information Fusion
T2  - Chinese Journal of Information Fusion
JF  - Chinese Journal of Information Fusion
VL  - 3
IS  - 3
SP  - 178
EP  - 188
DO  - 10.62762/CJIF.2025.472730
UR  - https://www.icck.org/article/abs/CJIF.2025.472730
KW  - multimodal sensor fusion
KW  - spatio-temporal information
KW  - intent prediction
KW  - dynamic visual modulation
KW  - virtual reality
KW  - human-in-the-loop systems
AB  - Virtual Reality (VR) systems require extremely low latency and accurate user intent prediction to ensure immersion. However, traditional visual modulation techniques primarily rely on single-sensor inputs (e.g., IMU angular velocity), which suffer from inherent mechanical delays and fail to capture complex spatial-temporal dynamics, ultimately leading to severe visually induced motion sickness (VIMS). Existing methods lack effective fusion of multi-source heterogeneous data, resulting in delayed responses and abrupt visual transitions. To address this, this study proposes a novel spatio-temporal information fusion framework for adaptive visual modulation in VR. The proposed architecture deeply fuses high-frequency eye-tracking signals (physiological intent), IMU data (physical kinematics), and dense optic flow (environmental spatial features). Specifically, a cross-modal temporal intent prediction model is established to fuse gaze and head-motion data, achieving early action anticipation. Simultaneously, a spatial data fusion module dynamically allocates asymmetric visual weights based on motion-risk mapping. Finally, a fusion-driven continuous state modulation solver, utilizing Bézier and Sigmoid functions, is introduced to eliminate multi-source asynchronous jitter and ensure C1 continuity. Experimental results demonstrate that the proposed multimodal fusion framework, by exploiting the known 80--120\,ms physiological precedence of eye movements over head rotation, enables approximately 100\,ms of anticipatory visual modulation, significantly outperforming baseline single-source methods. The approach yields a 27% reduction in Simulator Sickness Questionnaire (SSQ) scores, improves physiological comfort (an 18% increase in HRV), and drastically reduces mask perceptibility to 34%, all without sacrificing task performance. The findings provide an effective multimodal information fusion paradigm for human-in-the-loop immersive systems.
SN  - 2998-3371
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Shang2026Adaptive,
  author = {Wenli Shang and Kazlova Alena},
  title = {Adaptive Field-of-View Restriction Based on Sigmoid and Bézier Smoothing Control Functions},
  journal = {Chinese Journal of Information Fusion},
  year = {2026},
  volume = {3},
  number = {3},
  pages = {178-188},
  doi = {10.62762/CJIF.2025.472730},
  url = {https://www.icck.org/article/abs/CJIF.2025.472730},
  abstract = {Virtual Reality (VR) systems require extremely low latency and accurate user intent prediction to ensure immersion. However, traditional visual modulation techniques primarily rely on single-sensor inputs (e.g., IMU angular velocity), which suffer from inherent mechanical delays and fail to capture complex spatial-temporal dynamics, ultimately leading to severe visually induced motion sickness (VIMS). Existing methods lack effective fusion of multi-source heterogeneous data, resulting in delayed responses and abrupt visual transitions. To address this, this study proposes a novel spatio-temporal information fusion framework for adaptive visual modulation in VR. The proposed architecture deeply fuses high-frequency eye-tracking signals (physiological intent), IMU data (physical kinematics), and dense optic flow (environmental spatial features). Specifically, a cross-modal temporal intent prediction model is established to fuse gaze and head-motion data, achieving early action anticipation. Simultaneously, a spatial data fusion module dynamically allocates asymmetric visual weights based on motion-risk mapping. Finally, a fusion-driven continuous state modulation solver, utilizing Bézier and Sigmoid functions, is introduced to eliminate multi-source asynchronous jitter and ensure C1 continuity. Experimental results demonstrate that the proposed multimodal fusion framework, by exploiting the known 80--120\,ms physiological precedence of eye movements over head rotation, enables approximately 100\,ms of anticipatory visual modulation, significantly outperforming baseline single-source methods. The approach yields a 27\% reduction in Simulator Sickness Questionnaire (SSQ) scores, improves physiological comfort (an 18\% increase in HRV), and drastically reduces mask perceptibility to 34\%, all without sacrificing task performance. The findings provide an effective multimodal information fusion paradigm for human-in-the-loop immersive systems.},
  keywords = {multimodal sensor fusion, spatio-temporal information, intent prediction, dynamic visual modulation, virtual reality, human-in-the-loop systems},
  issn = {2998-3371},
  publisher = {Institute of Central Computation and Knowledge}
}