Automatically extracting performance data from recordings of trained singers.

Abstract

Recorded music offers a wealth of information for studying performance practice. This paper examines the challenges of automatically extracting performance information from audio recordings of the singing voice and discusses our technique for automatically extracting information such as note timings, intonation, mbrato rates, and dynamics. An experiment is also presented that focuses on the tuning of semitones in solo soprano performances of Schubert's "Ave Maria" by non-professional and professional singers. We found a small decrease in size of intervals with a leading tone function only in the non-professional group.This paper describes the challenges that arise when attempting to automatically extract pitchrelated performance data from recordings of the singing voice. The first section of the paper provides an overview of the history of analyzing recorded performances. The second section describes an algorithm for automatically extracting performance data from recordings of the singing voice where a score of the performance is available. The algorithm first identifies note onsets and offsets. Once the onsets and offsets have been determined, intonation, vibrato, and dynamic characteristics can be calculated for each note.The main experiment of the paper, described in the third section, is a study of intonation in solo vocal performance, where both the note onsets and offsets and fundamental frequency were estimated automatically. In the study, six non-professional and six professional sopranos performed Schubert's "Ave Maria" three times a cappella and three times with a recorded piano accompaniment. Our analysis of these recordings focused on the intonation of ascending and descending semitones. We found that the A-B^ intervals with a leading tone function were on average 8 cents smaller than the non-leading tone A-Bl,, but that their average size was approximately the same as the other semitones performed in the piece, regardless of intervallic direction or accompaniment. PREVIOUS WORK ON THE ANALYSIS OF RECORDED PERFORMANCESInterest in studying recorded performances dates back almost as far as the birth of recordable media, beginning with Dayton Miller's (1916) work on visualized pitch information in recordings with phonophotographic apparati. The psychologist Carl Seashore and colleagues at the University of Iowa also undertook extensive work in performance

A Longitudinal Study of Intonation in an a cappella Singing Quintet

“…Each singer in this study consistently tended towards equal temperament during and across rehearsals, as demonstrated by the pitch drift analysis. These results corroborate previous investigations conducted by Devaney et al 4 , showing no evidence of pitch drift in a Benedetti's three-part exercise, and contrast the findings from Howard, 5,6 observing pitch drift in a four-part piece, in which the chords were linked via a tied note in each case, composed for the study. These findings suggest that tuning in singing ensemble might depend on the specific melodic/harmonic characteristics of the piece performed as well as the individual singers and combination of singers performing.…”

“…Chord number significantly affected the tuning, suggesting, similar to the finding with the major thirds, that intonation of the minor thirds might be context-specific. The variability of the minor and major thirds based on the chord number, including examples of intonation close to both ET and just intonation, is in line with previous results found by Devaney et al 4 when investigating minor and major thirds in a three-part progression written by Benedetti. The consistency and distribution of the major and minor thirds' tuning, as quantified by the SD and range of the thirds' size respectively, did not differ pre-rehearsal and postrehearsal, and across rehearsals and repetitions, suggesting that this tuning behavior was highly repeatable in relation to the minor and major thirds and typified the ensemble.…”

“…Empirical research in this area, whilst sparse, has focused on different perspectives of choral tuning including predictions of pitch drift, pitch drift in performance, and perception preferences for different intonation systems. 3 Investigating tuning practices in a cappella part singing, Devaney et al 4 found no evidence of pitch drift in an exercise written by Benedetti in the sixteenth century to illustrate potential pitch drift associated with ''pure tuning,'' when performed by four expert 3-part ensembles. They hypothesized that this was due to the shortness of the exercise and the likelihood of retaining a pitch memory for the start of the piece throughout the eight-bar excerpt.…”

“…Analyzing interval sizes performed by four expert 3-part vocal ensembles, Devaney et al 4 found that whilst most vertical and horizontal tuning was in line with equal temperament both minor and major thirds varied in tuning with examples of intonation close to equal temperament, just intonation and Pythagorean tuning. Major 6ths were found to be consistently tuned to equal temperament.…”

Octave-generalized analysis of chord progressions: Diatonic/fifth relations, missing fundamentals, completion tones

“…Most Western tonal music is based on a chromatic scale that divides the octave into 12 approximately equal divisions. The present study is confined to that scale, understood to be approximately tuned-each scale degree plus/minus a few tens of cents (Devaney et al, 2011). We assume that intonation and tuning have always been variable and dependent on musical experience (cf.…”