TempEst: Harpsichord Temperament Estimation in a Semantic Web Environment

Abstract

Issues concerning tuning and temperament bear relevance to music research in areas such as historical musicology, performance and recording studies, and music perception. We have recently demonstrated that it is possible to classify keyboard temperament automatically from audio recordings of standard musical works to the extent of accurately distinguishing between six different temperaments often used in harpsichord recordings.

The current paper extends this work by combining digital signal processing with semantic computing and demonstrates the use of the temperament classifier in a Semantic Web environment. We present the Temperament Ontology which models the main concepts, relationships, and parameters of musical temperament, and facilitates the description and inference of various characteristics of specific temperaments. We then describe TempEst, a web application for temperament estimation. TempEst integrates the classifier with ontology-based information processing in order to provide an extensible online service, which reports the class and properties of both known and unknown temperaments. TempEst allows users to upload harpsichord recordings, and provides them with an estimated temperament as well as other inferred characteristics of the instrument's tuning.

Acknowledgements

The authors acknowledge the support of the School of Electronic Engineering and Computer Science, Queen Mary University of London, and the EPSRC-funded ICT project OMRAS2 (EP/E017614/1).

Notes

¹See Scala home page: http://www.huygens-fokker.org/scala

²OMRAS2 web site: http://www.omras2.org

³RDF is a data model primarily used on the Semantic Web for expressing simple statements in the form of subject-predicate-object. It has several different serialization formats such as XML/RDF and N3. See http://www.w3.org/TR/rdf-primer/ for an introduction.

⁴ http://purl.org/ontology/temperament

⁵Due to the dynamic nature of the system, further rules may be added to reflect accumulated experience.

⁶VamPy is available from http://vamp-plugins.org/vampy. html

⁷See www.python.org for details on Python, an increasingly popular language in the scientific community.

⁸A cent is one hundredth of a semitone, i.e. one twelve-hundredth of an octave.

⁹Polyphonic transcription is still regarded an unsolved problem (Casey et al., 2008; Klapuri, 2009).

¹⁰ http://purl.org/ontology/temperament

¹¹In this context: representing a relation by an object in order to reason about the relation itself.

¹²Vamp plugins: http://vamp-plugins.org/download.html

¹³See http://vamp-plugins.org/download.html#install

¹⁴Available from: http://sonicvisualiser.org/

¹⁵Available from: http://omras2.org/SonicAnnotator

¹⁶Available from: http://www.w3.org/2000/10/swap/doc/cwm.html

¹⁷Available at http://www.isophonics.net/sawa/

¹⁸Available at: http://www.cherrypy.org/

¹⁹Available at: http://www.isophonics.net/sawa/rec

²⁰Available at: http://www.isophonics.net/sawa/tempest

²¹The signal processing component of TempEst produces deviations from equal temperament. This description format is converted to the circle of fifths before inferring additional properties of the temperament, such as regularity.

²²The system is capable of fine-tuning the reference frequency within a vicinity of ± 40 cents, but requires a reference to avoid ambiguities between, for example, a B♭ in baroque pitch and an A in modern pitch, as both can potentially be associated with a frequency in the vicinity of 440 Hz.

²³For an explanation of RDF syntax and serialization formats including N3 see http://www.w3.org/TeamSubmission/turtle/ and http://www.w3.org/DesignIssues/Notation3.html. See also Fazekas et al. (2010) in the present issue for examples and applications in the music domain.