A Survey of Discrimination and Error Metrics for Spectrally Altered Instrument Tones

PhD Qualifying Examination


Title: "A Survey of Discrimination and Error Metrics for Spectrally Altered Instrument 
Tones"

Mr. Chung LEE


Abstract:

A musical instrument tone can be defined with four auditory attributes. They 
are pitch, duration, loudness and timbre. Pitch, duration and loudness are well 
defined and can be presented in a single continuum. However, timbre remains a 
poorly understood auditory attribute. Except its multidimensional nature, we 
understood very little about how human auditory system distinguish the timbres 
of different musical instruments. By Multidimensional Scaling (MDS) and 
studying the discrimination of altered tones, a number of acoustical attributes 
like spectral centroid, attack time and spectral variation were found to be 
important to timbre perception. Subjective evaluation of instrument tones 
always relies on listening tests, but they are sometimes impractical. 
Therefore, error metrics which can accurately predict the discrimination rate 
of altered tones are desirable. This survey gave an overview of a number of 
studies about timbre and introduced a number of previously suggested error 
metrics. We also suggested some further development on error metrics.


Date:     		Tuesday, 26 January 2010

Time:                   2:00pm - 4:00pm

Venue:                  Room 3501
 			lifts 25/26

Committee Members:      Prof. Andrew Horner (Supervisor)
 			Dr. Jogesh Muppala (Chairperson)
 			Dr. David Rossiter
 			Dr. Raymond Wong


**** ALL are Welcome ****