Séminaire n°12

Vowel Quality Extraction from Voice and Violins

Intervenant :  Robert Mores, Professor at the Mediatechnology Department of the University of Applied Sciences, Hambourg

Contact : mores(at)mt.haw-hamburg.de

Date : 08/02/10

Abstract : 
Vowel Quality Extraction from Voice and Violins Verbal sound descriptions generally refer to everyday life sounds or to other sensory experience. Quite often analogies between perceived sounds and the vocal tract are drawn. In this contribution, vowel quality is proposed as a compact and stable representation for sound descriptions. An automatic tool is developed for confidently extracting this feature from recorded sounds. When researching violin tones, almost every second sample reveals obvious vowel quality. Observation of vowel quality along musical lines reveals some of the perceived sound differences between violins - differences that can be directly imagined or even be imitated with one’s own vocal tract.

Séminaire n°13

Nasality in Voice and Musical Sounds

Intervenant :  Robert Mores

Date : 08/02/10

Abstract : 
'Nasality' is one of the attributes that is used often in sound quality assessments of musical instruments. Nasality is usually believed to be represented by a good portion of energy in a specific band around 1 kHz. Especially the strings community keeps referring to this band as the nasality band according to Dünnwald´s definition in the 1980s, with only little support from psychoacoustic tests however. Findings from two other research fields contradict Dünnwald´s definition. Clinical studies list several key ingredients to nasality, most of which are placed well outside the defined band. Moreover, the speech processing community investigates formant compositions including secondary formants rather than energy in the mentioned band. In general, a larger set of ingredients in a sound will add to the perception of nasality. This paper gives a brief survey on the different approaches to nasality. It also reviews the progress on automated feature extraction for this sound attribute.