Explanation of contents of MUMS revised

Explanation of contents of MUMS revised (database and installer)

1 Introduction

Sound databases are among the most relevant materials in Music Information Retrieval research, and one of the most problematic issues respect to the databases is the planning of their structure. The present work was inspired by the necessity of retrieving the contents of the McGill University Master Samples (Opolko & Wapnick, 2006) -MUMS- collection in an efficient fashion. As a side project of a timbre research (Eerola, Alluri, & Ferrer, 2008), the contents of the collection were thoroughly explored and analyzed (Eerola & Ferrer, 2008). During this process we realized a way to improve the structure and organization of the MUMS for our own purposes. The implementation of this re-organization is now an installer, which copies the samples from the original media into a new location. This document presents a brief explanation of the database design as well as issues related with the implementation of the installer.

2 Explanation of database fields

Remapped file names are composed of six fields that can be divided into two groups; first group refers to the location of the file in a hierarchical file-system structure, and the second group of fields contains a description of the sample. The names of the six fields are: Family, Subfamily, Type/Class, Attack/Playing style/Timbral specification, Sample type and Value (see fig.1). The structure of these fields will be explained next.



 Figure 1: Database structure 

2.1 Family

This is the first hierarchical level in the reorganization of samples. At this level, samples are grouped according to the well-known orchestration manual by Adler(1989). The categories at this level are: Bowed String, Brass, Definite Pitch Percussion, Indefinite Pitch Percussion, Keyboard, Plucked String, Woodwind. Percussion samples are divided in two groups: definite pitch and indefinite pitch.


2.2 Subfamily

As a second level in the hierarchy, subfamily field contains the generic name of the instrument that originated the sound sample. Each family is therefore subdivided in the following subfamilies:

  • Bowed String: acoustic bass, viol, viola, violin, violoncello.
  • Brass: cornet, french horn, treble cornett, trombone, trumpet, tuba.
  • Definite Pitch Percussion: crotales, glockenspiel, marimba, steel drum, timpani, tubular bells, vibraphone, xylophone.
  • Indefinite Pitch Percussion: membraphone, metal idiophone, multipercussion, wooden idiophone.
  • Keyboard: accordion, bass analog synth, celesta, harpsichord, organ, piano.
  • Plucked String: bass, guitar, harp, lute.
  • Woodwind: alto shawm, baroque recorder, bassoon, clarinet, contrabassoon, crumhorn, english horn, flute, oboe, pan flute, renaissance recorder, saxophone.


2.3 Type / Class

The typology at this level expresses an explicit definition of the instrument that produced the sample. In most cases it is very important to make this distinction, for example to establish the difference between acoustic and electric guitars. However, this distinction is not present in all cases. In order to provide consistency in the number of fields for better automatic retrieval, the expression "Standard" has been chosen to fill those places where no label was indicated in the original path & file name. The following instruments contain cases that differ from "Standard":

  • Bass: electric, fretless.
  • Clarinet: b flat, bass, contrabass, e flat.
  • Crumhorn: alto, bass, soprano, tenor.
  • Flute: alto, bass, piccolo.
  • Guitar: acoustic, electric.
  • Harpsichord: 8/4, 8 stop, buff stop.
  • Marimba: grand symphonic marimba.
  • Membraphone: 8in tom, 10in tom, 13in tom, 14in tom, 16in tom, 6in snare, 8in snare, bass drum, bongo, conga, cuica, kick 1, kick 2, kick 3, kick 4, kick 5, kick 6, kick 7, kick 8, marching snare, military drum, orchestral bass drum, power kick, processed tom, rock snare, snare, snare 2, snare 3, snare 4, snare 5, snare 6, timbale, timbales, tumba, tympani.
  • Metal Idiophone: 14in hi hat, 16 in cymbal, 18 in cymbal, 20 in cymbal, agogo bells, allpenglocken, brake drum, cabasa, cencerros, chinese cymbal, chinese gong, chinese rancancan, finger cymbal, flexatone, gong, hi hat, ice bells, large triangles, mark tree, orchestra cymbal, pandero, sizzle cymbal, sleigh bells, tambourine, triangle, turkish cymbal 15 in, turkish cymbal 20 in, waterphone.
  • Multipercussion: cha cha mix, compasa mix, conga tumba, cymbal snare crash, darbuka, egyptian mix, finger cymbal, hat cymbal kick, hat kick, mambo mix, pandero, rhumba mix, samba mix, shaker, triangle, triangle samba, triangle wawonko, wawonko mix, whistle.
  • Oboe: baroque, classical, oboe damore.
  • Organ: baroque plenum, brustwerk, crumhorn, dulzian, flutes, full organ, gemshorn, hauptwerk, koppelflote, nasat, pedals barplen, posaune, prinzipal, rauschpfeife, scharf, solo cornet, solo trumpet, symphonic plenum.
  • Pan flute: high, low, medium.
  • Piano: steinway, steinway concert hall grand, steinway hamburg grand.
  • Recorder: alto, soprano, tenor.
  • Renaissance recorder: alto, bassinet, great bass, quart, soprano, tenor.
  • Saxophone: alto, baritone, bass, soprano, tenor.
  • Trombone: alto, bass, tenor.
  • Trumpet: bach, in C.
  • Viol: bass, tenor, treble.
  • Violin: ensemble, violin1, violin2, violin3.
  • Wooden Idiophone: bamboo chimes, castanets, claves, log drum, maracas, ratchet, temple block, whip, wood block.


2.4 Attack / Playing style / Timbral specification

A substantial variety of timbral specific characteristics obtained within the same instrument exists in the collection. The information related to this specification is contained in the original file name. In some cases, these refer to an attack type, in other cases to a playing style or some other a timbral modification performed by digital means. Nevertheless this information is not always present, and in such cases we used the word: "Normal". Due to a variety of terms used in this field, we only list here those which have this special addition.

  • alto shawm standard
  • analog synth bass standard
  • baroque recorder alto
  • baroque recorder soprano
  • baroque recorder tenor
  • bassoon standard
  • celesta standard
  • clarinet b flat
  • clarinet bass
  • clarinet contrabass
  • clarinet e flat
  • contrabassoon standard
  • cornet standard
  • crotales standard
  • crumhorn alto
  • crumhorn bass
  • crumhorn soprano
  • crumhorn tenor
  • english horn standard
  • french horn standard
  • guitar acoustic
  • guitar electric
  • harp standard
  • harpsichord 8 4 stop
  • harpsichord 8 stop
  • harpsichord buff stop
  • lute archlute
  • lute renaissance 8 course
  • marimba
  • oboe baroque in c
  • oboe baroque with flattement
  • oboe classical
  • oboe oboe damore
  • oboe standard
  • organ baroque plenum
  • organ crumhorn
  • organ dulzian
  • organ gemshorn
  • organ koppelflote positive
  • organ nasat
  • organ solo trumpet
  • pan flute high
  • pan flute low
  • pan flute medium
  • renaissance recorder alto
  • renaissance recorder bassinet
  • renaissance recorder great bass
  • renaissance recorder quart
  • renaissance recorder soprano
  • renaissance recorder tenor
  • saxophone alto
  • saxophone baritone
  • saxophone bass
  • saxophone soprano
  • saxophone tenor
  • treble cornett standard
  • trombone alto
  • trombone bass
  • trombone tenor
  • trumpet bach
  • trumpet in c
  • tuba standard
  • tubular bells standard
  • viol bass
  • viol tenor
  • viol treble
  • violoncello standard
  • xylophone standard


2.5 Sample Type & Value

After the characteristics of the source had been expressed, the name of the file must describe the characteristics of the sample. This distinction is shown in the last two fields: the penultimate expressing if the type of event is a single sound or a pattern, and the last field represents the pitch in standard MIDI values. In the cases where the sample does not have a single pitch, or a description of an average pitch is not useful, the distinction made in the original file name was preserved. (see table. 1)


Type 5th. field      Meaning      

S Single event (stroke, hit, impulse, etc.)      
P Pattern / repetition of events      

Value 6th. field      Meaning      

0 Indefinite pitch      
21 - 114 MIDI value      
Wapnick Chord classification1      

Table 1: Sample type and value

2.6 General criteria

The remapping has been carried out trying to preserve as much as possible the original file name. However in order to give sound samples a name that could be effectively retrieved and/or discriminated by automatic means, plain spaces are substituted by a low dash character and the separation between fields is a regular dash. So for example the file named as: Bowed_String-Violin-Violin1-Non_Vibrato_Open-S-62.wav can be found at Bowed_String directory and then Violin and Violin1 sub-directories.


2.7 Special cases

  • Martelé has been written in the database without the accent to assure compatibility with different file coding systems.
  • Keyboard-Organ-Scharf-4_28_positif The relation between the aural, F0 pitch estimation, and original label could not be established, for that reason the original pitch name is kept in the new mapping as reference.
  • Bowed_String-Violin-Standard-Vibrato_Open These samples are not really sounding in vibrato style. (Sample Numbers: 5071, 5089, 5094)
  • Bowed_String-Acoustic_Bass-Standard-Plucked_Stopped-S-38 This sample contains a corrupt header. (Sample Number: 4039)
  • 'DVD 3/WOODWINDS/HISTORICAL WIND INSTRUMENTS/RECORDERS/Renaissance SOPRANO Recorder/RENAISS SOPRANO REC_D#6.wav' and 'DVD 3/WOODWINDS/HISTORICAL WIND INSTRUMENTS/RECORDERS/Renaissance SOPRANO Recorder/RENAISS SOPRANO REC_E6.wav' are a copy of the same sample (Sample Numbers: 6184, 6187). Only E6 (Woodind-Renaissance_Recorder-Soprano-Normal-S-88) is copied into the new location.
  • In the next cases, the Sample Type & Value part of the original typology has been preserved:
    • Bowed_String-Acoustic_Bass-Standard-Slide
    • Bowed_String-Acoustic_Bass-Standard-Improvised_Lines
    • Keyboard-Accordion-Standard-Chord
    • String-Guitar-Electric-Chord


3 Installers

The renaming procedure has been implemented in the form of an installer, a script that looks for the files as they were originally named in the collection, and makes a copy of them in the local hard-drive of the user, creating a new directory structure in the process. Currently, there is one version tested in Mac Os X 10.4.10), which can be easily edited to be used in any UNIX type systems. This installer is freely available from our website2. For Microsoft Windows version, the installer is not yet available but may be developed if need arises. For using the installer, you have to own a copy of the MUMS DVD3 and possess enough hard-drive space to store the new database (9.3 Gb).


1In relation to the classification of the chords, the original typology has been preserved. Similarly, special patterns such as Bass slides and improvised lines have been preserved in their original formulation.







   Adler, S. (1989). The study of orchestration (2nd ed.). New York, USA: Norton & Norton Company.

   Eerola, T., Alluri, V., & Ferrer, R. (2008). Emotional connotations of isolated instruments sounds. In Proceedings of the ICMPC10 conference.

   Eerola, T., & Ferrer, R. (2008). Instrument library (MUMS) revised. Music Perception, 25(3), 253-255. Available from http://caliber.ucpress.net/doi/abs/10.1525/mp.2008.25.3.253

   Opolko, F., & Wapnick, J. (2006). The McGill University Master Samples Collection on DVD (3 DVDs). Quebec, Canada: McGill University. Available from http://www.music.mcgill.ca/resources/mums/html/


Go to Download Form