|Budget Amount *help
¥2,600,000 (Direct Cost : ¥2,600,000)
Fiscal Year 2001 : ¥600,000 (Direct Cost : ¥600,000)
Fiscal Year 2000 : ¥500,000 (Direct Cost : ¥500,000)
Fiscal Year 1999 : ¥900,000 (Direct Cost : ¥900,000)
Fiscal Year 1998 : ¥600,000 (Direct Cost : ¥600,000)
The objective of this research project has been to develop necessary techniques required for "Stream Segregation" of tonal music in particular. Stream Segregation means to separate sounds from different sources and identify the sources, or to realize the same functions on a computer. Although general stream segregation is an integrated consequence of perceiving timbre, tone height, time information, direction perception and so forth, this project investigates possibility of computational stream segregation related to music. Concretely speaking, it treats separation of voice parts, tone name identification, key determination and structural analysis of individual musical excerpts for automatic scoring.
The final objective of this research project is to establish a confident perspective for realizing an automatic scoring system of ensembles, for example string quartets in particular. We reached the level to be sure to be able to develop a practical system at the end of the project, though
the final results are just finished as a graduation thesis and have not been published yet. Following is a brief description of the results obtained in the 4 year term of the project.
I. Tone height and Pitch
Perceptual height for tones presented sequentially is investigated, Cepstral Prominence Factor is proposed for representing easiness of pitch perception, and improvement of phonemic identifiability by vibrato for high pitch singing voices.
II. Perception of Pitch and Direction
Possibility of realizing a stereophonic system using the precedence effect is investigated, and obtained is that pitch perception precedes direction perception in the auditory system from perceptual experiments using sounds of missing fundamentals.
III. Implementation of the Theory of Harmony
Developed are "Bass Donnee System (BDS)" that enumerates all the allowable upper 3 voice solutions for given bass sequences by depth-first search. BDS is extended to "Soprano Donnee System (SDS) ." applicable to soprano tasks.
IV. Music in General
Developed are algorithms for determining the keys and structural analysis. Also discussed are standards of aesthetics from a statistical point of view and how to explain music after the collapse of tonality.
V. Automatic Scoring
Starting from identification of tone names of six notes simultaneously played on guitars, automatic scoring of string quartet comes into sight based on the "Multiple Harmonic Structure Model" developed in the project. Less