Exploring the perceived harshness of cello sounds by morphing and synthesis techniques - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Journal of the Acoustical Society of America Année : 2017

Exploring the perceived harshness of cello sounds by morphing and synthesis techniques

Résumé

Cello bowing requires a very fine control of the musicians' gestures to ensure the quality of the perceived sound. When the interaction between the bow hair and the string is optimal, the sound is perceived as broad and round. On the other hand, when the gestural control becomes more approximate, the sound quality deteriorates and often becomes harsh, shrill, and quavering. In this study, such a timbre degradation, often described by French cellists as harshness (décharnement), is investigated from both signal and perceptual perspectives. Harsh sounds were obtained from experienced cellists subjected to a postural constraint. A signal approach based on Gabor masks enabled us to capture the main dissimilarities between round and harsh sounds. Two complementary methods perceptually validated these signal features: First, a predictive regression model of the perceived harshness was built from sound continua obtained by a morphing technique. Next, the signal structures identified by the model were validated within a perceptual timbre space, obtained by multidimensional scaling analysis on pairs of synthesized stimuli controlled in harshness. The results revealed that the perceived harshness was due to a combination between a more chaotic harmonic behavior, a formantic emergence, and a weaker attack slope.
Fichier non déposé

Dates et versions

hal-01586525 , version 1 (13-09-2017)

Identifiants

Citer

Jocelyn Roze, Mitsuko Aramaki, Richard Kronland-Martinet, Sølvi Ystad. Exploring the perceived harshness of cello sounds by morphing and synthesis techniques. Journal of the Acoustical Society of America, 2017, 141 (3), pp.2121--2136. ⟨10.1121/1.4978522⟩. ⟨hal-01586525⟩
69 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More