GB2297859A - An apparatus for automatically generating music from a musical score - Google Patents

Abstract

The apparatus comprises a sensing device 2 to scan a musical score, and apparatus 3 to recognise the symbols, notation and other musical indications contained therein. This generates signals which, after suitable coding 5, are stored for subsequent processing. These stored codes may be adjusted 11 to provide an interpretation of the score in line with individual tastes and preferences. These codes are then processed 7 in combination with wave-form data 8 to generate a signal 9 to drive audio amplifiers so that the interpretation can be heard 10 and also recorded 12 if so desired.

Description

Patent Application AUTOMATIC SOUND GENERATION This invention relates to the automatic rendition of sounds derived from a musical score.

Conventionally, the performance of a composers work has been achieved by instrumentalists (and vocalists) reading from a written or printed score, with additional interpretation and guidance from a conductor, band-leader, arranger or similar person. Recording the performance requires the use of transducers and associated equipment.

This is a lengthy and expensive process, requiring much training and rehearsal of the instrumentalists to obtain a satisfactory performance. In addition, costly facilities and equipment are needed to record the final interpretation.

According to the present invention, there is provided automatic equipment to generate sound derived from the indications in a printed or manuscript musical score, to adjust this sound, if so desired, in line with the taste and requirements of a listener, and to record it in appropriate form, comprising of a mark sensing device to scan the score, apparatus to convert the signal from this to a coded representation, a device to store this representation permanently or semi-permanently, equipment to allow this representation to be amended and the altered version also to be stored, and apparatus to prepare and store a signal derived from such representations to drive devices used for play-back or recording of the sound.

A specific embodiment of the invention will now be described by way of example. Reference will be made to the accompanying drawing, Fig. 1. which shows the overall arrangement in broad detail The initial step is to use an optical scanner (2) on the written or printed score (1), to produce a digital image of it stored in a memory array. If appropriate, for example with a lengthy piece, the score may be scanned in sections and re-combined later, for ease of processing.

Automatic Sound Generation The array is then processed, using character recognition procedures (3), to extract information contained in the score. This has four aspects. One is harmonic, the recognition of the stave itself, and basic symbols entered on it such as clef signs, bar lines, sharps and flats, repeats, and notes and rests of various types together with tie signs. It is also necessary to recognise the number of parts shown in the score.

The second is the recognition of symbols representing musical dynamics, entered around (or occasionally in) the stave; these usually take the form of words or abbreviations, such as 'mf', 'pp', 'dim', crews.', though symbols may be used rather than words, particularly for glissandi and similar dynamic effects.

The third aspect covers tonal qualities. The primary means of recognising these is from indications of the instrument to be associated with each part (e.g. 'violin', 'piano', 'trumpet', etc.) which, if present, can be sensed from the score. A secondary source is markings to indicate specific instrumental effects, such as 'pizzicato', 'vibrato', 'mute', rasgueado', etc.

Finally, the fourth aspect is indications covering the whole, or a substantial part of, the score, such as key, time signature, tempo and style. The key may be shown separately (e.g. 'Key=F') or merely by the key-signature on the stave. The tempo may be shown by words such as 'lento', 'presto' or by a time-value for a reference note (e.g.

crotchet =112). Style will be indicated by words such as 'Con moto', 'lyrically', 'syncopated', 'rumba' etc. This indicates the relative stress on each beat of the bar, and the degree of musical 'attack' when sounding notes.

The number and range of symbols (4) that need to be recognised and interpreted under these headings depends upon the intended use. For simplistic renderings of the score, only comparatively few are needed.

More sophisticated interpretations, or more accurate and faithful interpretations, will demand a greater range of symbols to be recognised and interpreted.

Automatic Sound Generation Information sensed from the score under these four headings is coded (5) into a digital representation and compiled into four corresponding, associated, files (6).

The first is the harmonic file. The preliminary step here is to recognise how many musical parts are shown. This is achieved by counting how many staves are joined together vertically by lines at either or both sides of the page.

Each part is then treated separately, and processed to yield data on the relative pitch and duration of each note. This is derived from the shape (which indicates relative duration), and clef sign plus position on the stave (which indicates pitch) of individual notes and rests. Allowance is made for tie signs, and key signature respectively.

Pitch information is also adjusted to allow for any accidentals shown.

Where chords are shown, these are treated as the simultaneous sounding of the constituent notes.

Any 'repeat' signs are used to duplicate the appropriate information and append it to the file, applying corrections to allow for alternative endings to repeats shown in the score.

Where the score only gives a partial indication of what is to be played, as in bass continuo parts in some early music, pitch and duration information is filled in through reference to a file of standard sound patterns.

Thus the 'Harmonic' file will consist of a series of codes to represent relative pitch and duration of each note, silence, or chord, compiled separately for each part. It may also include data on the position of bar lines, to assist in subsequent synchronisation of the four files, and to ease the making of any desired amendments.

The Dynamics file will also be compiled separately for each part.

This will comprise codes for the loudness of each bar, the rate of change of loudness, and the bar(s) to which these refer. Where no dynamic indications are given on the score, default values will be assigned.

Automatic Sound Generation The Tonal file, again referring to each separate part, will contain a code for the overall tonal characteristics, plus codes for special instrumental effects and the bars (or individual notes) to which these apply. Again, where no tonal indications have been detected, default values will be supplied.

Vocal parts represent a special case here. If the listener is not concerned about the actual lyric, then these are allocated tonal characteristics as for any other instrument. If the lyric is thought to be important, special provision can be made to scan it and create a file for it, to be processed later through speech synthesis devices and subsequently incorporated in the overall file.

If the score shows a transposing instrument, such as a clarinet, for a part, information about this is recorded in the file so that appropriate adjustments to the pitch information can be made.

The Style file contains codes for the time-signature, an absolute value for the duration of a reference note, a value for the pitch of a standard reference note, the key signature, the style, plus any changes to these in the course of the piece and the time or bar when such changes occur. Optionally, it may include statistics about the piece, such as number of bars, playing time, etc.

These files may be transferred to a storage device such as a floppy disc, if so desired, for use at a later date.

Subsequently, these files are processed further by combining them (7), either to generate a signal suitable for driving an amplifier or by compiling a file suitable for use by sound sequencer and synthesiser devices.

The principal feature of this further processing is the allocation of tonal characteristics to the harmonic information, controlled by data in the Tonal file. This data is used to select wave envelopes and/or waveforms from a reference file (8) and associate them with the pitch and duration information in the Harmonic file. These tonal qualities may be approximations of various degrees of accuracy to Automatic Sound Generation the sounds of real instruments or other natural sounds, or be synthetic creations.

Another process here is to convert relative pitch and duration values to absolute ones, by using the reference values in the Style file. Basic information on loudness of each sound is acquired from the Dynamics file, then modified by data from the Style file to allow for the relative emphasis on different beats of the bar. Wave envelopes are modified in accordance with data on the required 'attack' from the Style file. At this stage, stereo, quadraphonic or other acoustic effects can be incorporated, by controlling the level of signal passed to each of several channels.

This unified version provides the basis for generating a signal (9) suited to the needs of the playing equipment available. This may be, for example, an electrical analogue signal for an amplifier, or a digital file for sound synthesis circuits and devices.

On hearing the interpretation thus produced (10), a listener can decide whether he wishes to adjust it in any way. Equipment (11) is provided to allow him to do this by accessing any or all of the Harmonic, Dynamics, Tonal or Style files, altering the necessary parameters, and storing the new versions of these files. These can then be passed through the processing devices again, to permit him to listen to and judge the amended interpretation. This cycle may be repeated as often as desired.

Once a satisfactory interpretation has been arrived at, it can then be transferred (12) to a conventional recording medium, such as cassette tape.

Claims (6)

Automatic Sound Generation Claims

1. Automatic sound generation equipment to provide an interpretation and/or a recording of a musical score directly from the page, without the intervention of instrumentalists or vocalists.

2. Automatic sound generation equipment as claimed in claim 1 wherein mark sensing devices adapted to the recognition of musical symbols, notation and associated terms from a score are provided.

3. Automatic sound generation equipment as claimed in claim 1 or claim 2 wherein signals are generated in accordance with the indications in a musical score.

4. Automatic sound generation equipment as claimed in claim 1 or claim 2 wherein apparatus is provided to generate codes from the output signals of a mark sensing device.

5. Automatic sound generation equipment as claimed in claim 1 or claim 4 wherein is provided means to store codes representing musical values.

6. Automatic sound generation equipment as claimed in claim 1 or claim 4 or claim 5 wherein is provided equipment to alter or adjust coded representations of music to produce an interpretation in line with individual tastes and preferences.

6. Automatic sound generation equipment as claimed in claim 1 or claim 3 wherein is provided a device to combine codes representing separate musical values and tonal information, and hence to generate a signal suitable for driving devices for the play-back or recording of sound.