AI & ML
EE-Motion: An EEG-based Web Architecture for Live Affective Music Recommendation
arXiv cs.HC0 views
arXiv:2611.21032v2 Announce Type: replace-cross Abstract: Music listening is intrinsically tied to emotion, which drifts continuously within a single session. Although modern browsers interface with wearable brain-computer interfaces, recommendation systems remain structured around discrete logs and lack the infrastructure to integrate continuous physiological signals. Furthermore, existing affect-aware models rely on static user profiles and offline training, failing to serve anonymous newcomers without model retraining. We present a web platform that resolves both challenges. Pairing with a consumer electroencephalogram (EEG) headband over Web Bluetooth, the system captures live EEG, mapping frontal alpha asymmetry and arousal index to normalized valence-arousal proxy coordinates on Russell's circumplex plane. Driven by these coordinates, an affective adapter unifies structurally diverse recommendation backbones under a shared live scoring mechanism. By parameterizing anonymous visitors through a few favorite tracks, the module recommends tracks conditioned on live proxy coordinates. This demonstrates that real-time EEG-derived control signals can steer recommendations without persistent user tables or model retraining.
Source:https://arxiv.org/abs/2611.21032PreviousA 1000-hour EEG-EMG-audio dataset of Japanese speech productionNextA polynomial-time decoding algorithm up to half the square-root bound for quadratic residue and generalized quadratic residue codes
Related posts
DexFormer 2.0: Towards a Cross-Embodiment Foundation Policy for Dexterous Grasping
0arXiv cs.RO
RoboKit: Accelerating Parallel Differentiable Optimization for Robot Learning
0arXiv cs.RO
Cleanly-DINO: Efficiently Learning Visual Task Spaces for Robotics
0arXiv cs.RO
SAFER: Navigating the Exploration-Exploitation Dilemma in World-Model-Based Planning for Autonomous Driving
0arXiv cs.RO
PlanDiv: Test-Time Scaling with Latent Plans for Vision-Language-Action Models
0arXiv cs.RO
MoViC: Real-Time Humanoid Whole-Body Control from Monocular Video Streams
0arXiv cs.RO