<ul><li>Introduction of an unsupervised framework for detecting audio patterns in musical samples (loops) through anomaly detection techniques to address challenges in music information retrieval.</li><li>Combination of deep feature extraction with unsupervised anomaly detection using a pre-trained Hierarchical Token-semantic Audio Transformer (HTS-AT) and Feature Fusion Mechanism (FFM).</li><li>Utilization of one-class Deep Support Vector Data Description (Deep SVDD) to learn normative audio patterns by mapping them to a compact latent hypersphere, showing improved anomaly separation in evaluations on curated bass and guitar datasets.</li><li>Research presents a fully unsupervised solution for processing diverse audio samples, enabling effective pattern identification through distance-based latent space scoring, overcoming previous limitations.</li></ul>

Learning Normal Patterns in Musical Loops

Discover more