Unsupervised discovery of acoustic patterns in bird vocalisations employing DTW and clustering

This paper presents a method for an unsupervised discovery of acoustic patterns in bird vocalisations recorded in real world natural environments. The proposed method employs sinusoidal detection to provide frequency tracks which are used as features to characterise bird tonal vocalisations. A variant of dynamic time warping, capable of searching for multiple partial matchings, is used to segment the data based on these frequency track sequences. Agglomerative hierarchical clustering approach is then employed to cluster recurring segments. Evaluations are performed on audio recordings provided by the Borror Laboratory of Bioacoustics. The obtained results indicate that structurally distinct stereotyped acoustic units can be determined.