MDS of instrument sounds (Christian Kaernbach )

Subject: MDS of instrument sounds
From:    Christian Kaernbach  <chris(at)PSYCHOLOGIE.UNI-LEIPZIG.DE>
Date:    Sat, 26 May 2001 13:53:36 +0200

Thanks to all who responded to my inquiry on sound databases of musical instruments. In my next mail I will summarize the responses. I plan to do a course room experiment on MDS of musical timbres on Monday afternoon. It will be small (only four sounds) and informal, but I want to do it right. If anybody is willing to listen to my sounds and to give me advice: See postscript. - Christian PS: I have prepared four sounds (piano, clavichord, organ, violin). You can listen to them at Details of sound generation see file readme.txt. Due to technical reasons (available CDs), the sounds are rather short (610 ms, with only about 200 ms full sound for organ and violin), but that should only make them more comparable and focus attention to timbre (I hope so). I plan to first demonstrate the four sounds, two or three times each. Then I will present the six possible combinations, as sound files that contain the two 610-ms sounds, separated by 500 ms silence, see I will first cycle two times presenting all pairs. During the third and last cycle students should mark their ratings on a 9-point (1...9) dissimilar-similar scale (neutral = 5). Then I will restart to present the four single sounds, and ask the students to give their ratings on the following scales (all 9-point 1...9, neutral = 5): Klang hell/scharf/brillant --- dumpf/weich/dunkel (timbre bright/strident/brillant --- dull/soft/dark) Toneinsatz weich/sanft --- hart/abrupt/schlaghaft (attack soft/gentle --- hard?/abrupt/beating) Klang rauh/uneben --- glatt/gleichmäßig (sound rough/harsh/bumby --- smooth/even) unangenehm --- angenehm (unpleasant --- pleasant) I would be very happy to hear opinions as to whether these scales are appropriate guesses of what could be similarity dimensions for those four sounds. Especially if you have ideas which dimensions I did not ask for but are more plausible candidate dimensions, or which of those dimensions I ask for are not appropriate at all and should be skipped, please let me know. Something that I should do/ not do? Anything else that I should consider?

This message came from the mail archive
maintained by:
DAn Ellis <>
Electrical Engineering Dept., Columbia University