Prediction of music pairwise preferences from facial expressions
summary
Predicts pairwise music preferences from webcam-recorded facial expressions using an Affectiva-like API that returns high- and low-level signals, trains Random Forest and Gradient Boosting on delta features from 15s snippets, and outperforms a listening-time baseline (accuracy and RMSE improvements reported; classification accuracy rises noticeably above majority and time-based baselines). Study uses 75 subjects, 200 unknown clips, 10 Hz feature capture, and finds facial features vs listening time, delta contempt valence joy sadness, random forest / gradient boosting.
AI-written summary · about
chapters (12)
- 0:00Introduction to talk
- 0:26Background on recommender systems
- 1:53Types of preferences
- 3:01Emotional responses to music
- 4:46Research questions and goals
- 5:55User study design
- 8:11Data collection from facial expressions
- 9:35Feature extraction and modeling
- 11:03Results of prediction models
- 12:52Personality traits impact on predictions
- 13:55Conclusion and future work
- 14:57Q&A session
related talks
49mAffective Computing – Bringing Humans and Machines Closer Through Emotions
15mRecognizing Emotional Expression in Game Streams
59mTechnical Workshop: Sarcastically Speaking: Unlocking Multi-modal Sentiment Analysis With NLP and Facial Expressions
34mEnhancing Customer Experiences with Emotional AI,
07mAssessing Cognitive Performance Using Physiological and Facial Features: Generalizing Across ...
31mOff-the-shelf HuggingFace models for audio deepfake detection
50mAffective Computing
15mC-Face: Continuously reconstructing facial expressions by deep learning contours of the face ...