Title: A Multimodal Dataset Combining Photoplethysmography and Continuous Behavioral Responses to ASMR and Nature Videos

Abstract:

Autonomous Sensory Meridian Response (ASMR) is a somatosensory experience marked by pleasant tingling sensations and a reduction in cardiovascular activity. Despite its distinct characteristics, the advancement of ASMR research has been constrained by a significant shortage of standardized, publicly available multimodal datasets. To overcome this barrier, we introduce REST-ASMR (Response to Environmental & Sensory Triggers), a synchronized multimodal dataset engineered to record both behavioral reports and physiological changes during ASMR induction, utilizing nature-relaxation videos as control stimuli.

The dataset comprises high-resolution photoplethysmography (PPG) data, time-aligned audiovisual content, and continuous subjective annotations collected from 34 participants. Our technical validation demonstrated high stimulus efficacy, evidenced by a 97% responder rate. We observed significant stimulus-specific inter-subject agreement (p < 0.05) alongside robust, ASMR-specific cardiovascular deceleration derived from PPG signals.

Furthermore, a Bidirectional Long-Short Term Memory (Bi-LSTM) model was employed to predict subjective ASMR tingle states. Under a rigorous, leakage-free subject-video double-independent 4-fold cross-validation framework, the model achieved perfect accuracy in classifying videos at the ASMR versus Nature level. It also attained a frame-level global mean accuracy of 75.51%, a macro F1-score of 71.86%, and 100% specificity for the Nature baseline. REST-ASMR provides a dense temporal foundation for affective computing, multimodal research, and the creation of personalized models for relaxation-related responses.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC