Some general information on flamenco and the topics we are working on…
Flamenco singing: A brief introduction
Flamenco is an eminently individual yet highly structured form of music. Improvisation and spontaneity play a central role, but both heavily lean on an extremely stable organization of the musical material. Flamenco music has developed by coalescence of several music traditions into a rich melting pot, whose combination of singing, dancing and guitar playing is distinctive. Apart from the influences of the Jews and Arabs, flamenco music shows the imprint of the culture of the Andalusian Gypsies, who decisively contributed to its form today.
- Instability of pitch. In general, notes are not clearly attacked. Pitch glides or portamenti are very common.
- Sudden changes in volume (loudness). Those sudden changes are very often used as an expressive resource.
- Short melodic pitch range. It is normally limited to an octave and characterized by the insistence on a note and those contiguous to it.
- Intelligibility of voices. Lyrics are important in flamenco, and intelligibility is then desirable. For that reason, contralto, tenor, and baritone are the preferred voice tessituras.
- Timbre.Timbre characteristics of flamenco singers depend on the particular singers. As relevant timbre aspects, we can mention breathiness in the voice and absence of high frequency (singer) formants.
These characteristics contrast with classical singing styles, where precise tuning and timing are important, and where timbre is characterized by stability, absence of breathiness, and high-frequency formants (i.e., the singer formant).
Flamenco a cappella cantes
A cappella cantes constitute an important group of styles in flamenco music. They are songs without instrumentation, or in some cases with some percussion. Examples of a cappella styles are tonás, deblas, martinetes, carceleras, nanas, saetas, and some labor songs.
From a musical point of view, a cappella cantes retain the following properties:
- Conjunct degrees. Melodic movement mostly occurs by conjunct degrees.
- Scales. Certain scales such as the Phrygian and Ionian mode are predominant. In the case of the Phrygian mode, chromatic rising of the third and seventh degrees is frequent.
- Ornamentation. There is also a high degree of complex ornamentation, melismas being one of the most significant devices of expressivity.
- Microtonality. Use of intervals smaller than the equal-tempered semitones of Western classical music.
These features are not exclusive to a cappella cantes and can be found to various degrees in other flamenco styles.
The classification of flamenco cantes in general and of a cappella cantes in particular is subject to many difficulties, and such a classification is not yet clearly established in the flamenco literature. Two cantes belonging to the same style may sound very different to an unaccustomed ear. In general, underlying each cante there is a melodic skeleton. This melodic skeleton is filled in by the singer by using different kinds of melismas, ornamentation and other expressive resources. An aficionado’s ears recognize the wheat from the chaff when listening, and appreciate a particular performance in terms of the quality of the melodic filling, among other features. A flamenco aficionado recognizes both versions as the same cante because certain notes appear in a certain order (they are called main notes). What happens between two of those notes does not matter regarding style classification, but does matter for assessing a performance or the piece itself. As an example we include here the first phrase of a Debla (same lyrics) performed by Antonio Mairena and Chano Lobato.
Automatic transcription of flamenco singing
Why should we transcribe flamenco singing?
Unlike many other genres, flamenco is an orally transmitted art form: Songs were passed from generation to generation without having ever been written down in form of a musical score. Only a few manual transcriptions exist. Since flamenco performances are to the largest extend improvisational, automatic transcriptions are of minor importance for practicing musicians.
But: For music information retrieval (MIR) tasks, a compact but accurate melody representation is crucial for a number of related tasks, i.e. melodic similarity computation, pattern detection or style classification. The development for such tools sets the basis for user-oriented applications (how about “query by cante?”) as well as large-scale musicological studies.read more
Why is it so difficult?
Extracting a note transcription from an audio recording is considered one of the most challenging tasks in MIR. Flamenco singing is particularly difficult for two reasons: It is flamenco and it is singing.
A major difficulty of automatic transcription is the segmentation of a pitch contour into single notes. The singing voice is a particular challenging instrument to transcribe, since it is pitch-continuous and non-percussive. In flamenco music, we furthermore encounter a large amount of fast and unstable pitch fluctuations, i.e. vibrato, glissandi, micro-tonal ornamentations, melisma or tuning instabilities. Also, in most flamenco recordings the singing voice is accompanied a guitar, which poses an additional challenge.
How does it work?
An in-depth technical description of the CANTE transcription system can be found here.
The process involves five main stages:
1.) Channel selection: In most stereo flamenco recordings the voice is more dominant on one of the two channels. We use a set of low-level descriptors to select this channel for further processing.
2.) Pitch extraction: We use a predominant melody extraction algorithm to extract the fundamental frequency corresponding to the main melodic line. In the case of flamenco, this corresponds to the singing voice and some sections where the guitar takes over the main melody, i.e. the falsettas.
3.) Contour filtering: In this stage, part of the contour which belong to the guitar are eliminated based on low-level descriptors.
4.) Note segmentation: Now the continuous fundamental frequency contour is segmented into single notes. This is done based on contour and volume characteristics.
5.) Pitch quantisation: Once the segments are found, a MIDI pitch value is assigned to ever segment. Since the pitch contour shows fast fluctuations and the tuning might not be exact, we estimate the tuning and also analyse the overall pitch statistics of the entire song.
How does it sound like?
Here are some examples:
audio + transcription
audio + transcription