Abstract
This publication describes a media recommendation system that conditions content selection on a listener’s temporal physiological state rather than on external context alone. A biometric-sensing wearable device streams heart rate (HR), heart-rate variability (HRV), and inertial (e.g., accelerometer or gyroscope) data to a host media device, which buffers the multi-stream time series over a rolling look-back window. The buffered series is decomposed into a personalized seasonal component (circadian rhythms, baseline metabolic variation, historical resting-heart-rate zones, etc.) and a trend component that outputs the velocity and acceleration of the physiological state. A context classifier fuses biometric and motion signals to classify an elevated heart rate as physical exertion or cognitive stress. A sequential time-series model then forecasts the expected physiological condition at the next content boundary (e.g., the next song), and the recommendation engine selects media aligned with that forecast trajectory. User feedback signals such as skips, likes, and playlist additions serve as rewards in a reinforcement-learning model that continually improves the recommendation algorithm. The approach reduces skip rates and user friction caused by recommendations that conflict with biological and psychological needs of the user.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Weinmeister, Karl Joseph, "Adaptive Media Content Recommendation Based on Temporal Physiological States", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11701