Multimodal Joint Head Orientation Estimation in Interacting Groups via Proxemics and Interaction ...
summary
A shared LSTM social context pooling model jointly regresses head orientations for all members of a group using 10‑second sequences of wearable neck badges (speaking status, tri‑axial acceleration) plus room‑camera body positions and orientations; the group hidden states are combined into a group context vector each timestep. The system replaces coarse angle bins with regression over classification, matches or outperforms a rule‑based listener‑toward‑speaker baseline and frame‑based and simplified sequence baselines, generalizes to an unseen event dataset, and adds upper‑body keypoints and group annotations to the public corpus.
AI-written summary · about
related talks
09mUsing Gaze Behavior and Head Orientation for Implicit Identification in Virtual Reality
41mSocial Behavior Prediction from Video Observations
06mWeakly Supervised Multi-Task Representation Learning for Human Activity Analysis Using Wearables
10mHAGI: Head-Assisted Gaze Imputation for Mobile Eye Trackers
14mAssisting group activity analysis through hand detection and identification in multiple ...
47mHow a Tiny Head Detector Powers Intelligent Video Conferencing
10mREGROUP: A Robot-Centric Group Detection and Tracking System
10mProactive Sensing for Improving Hand Pose Estimation