Speaker
Description
The underrepresentation of non-western musical traditions in open-source datasets poses a persistent challenge for computational psychoacoustic analysis, particularly due to low-quality recordings or insufficient sample sizes that characterize these collections. This study examines the use of statistical inference methods for the psychoacoustic characterization of non-western traditional percussion instruments under conditions of small sample size. The method is applied to three playing techniques of the Colombian Alegre drum — Abierto (Tone), Bajo (Bass), and Quemado (Slap)— analyzed through psychoacoustic features. Recordings were obtained in an anechoic chamber using an artificial head system. To address the limited dataset size (N = 17), the bootstrap method is applied to generate 95% confidence intervals, and Hedge's g is computed to assess effect sizes, enabling a reliable and robust comparison of timbral profiles across techniques. Preliminary results reveal distinct timbral profiles for each technique, with the Bajo exhibiting the lowest centroid and widest bandwidth, the Quemado the highest roughness and spectral centroid, reflecting weak fundamental energy and broad spectral distribution up to 5kHz, and the Abierto the highest crest factor. Timbral profiles for each technique are visualized through radar charts, allowing a direct comparison of psychoacoustic features across playing techniques. These findings demonstrate that meaningful psychoacoustic characterization is achievable with small datasets, supporting the documentation and digital preservation of non-western musical heritage.