Home | Biodata | Biography | Photo Gallery | Publications | Tributes
[Back to Music and Musical Instruments List]

Music and Musical Instruments


Computer Translation of South Indian Music into Staff Notation
Proceedings of the International Symposium on Microwaves and Communication
IIT Kharagpur, pp 170-173, 29-30, December 1981
Gift Siromoney, R.Chandrasekaran and M.Chandrasekaran

This paper is concerned with machine recognition of musical notation and makes use of computer methods presented by the authors to translate South Indian musical notation in machine printed or hand-printed forms into western staff notation.

In the last two decades a large number of research projects linking the latest computer technology with music analysis, printed reproduction and composition have been proposed and implemented with varying degrees of success. In all of them, the major consideration has been to develop a convenient method for translating a music manuscript into machine usable form. The problem associated with the translation is that of automatic recognition of music data in printed or handwritten form. Once a computer program that can recognize music notation in either of the two forms is developed then a number of problems can be automated like the translation of music data into any desired key and so on.

David Prerau1 has developed a computer method for automatic recognition of printed musical scores in western notation. The notes recognized by the computer were put in the Ford-Columbia notation also called DARMS. An automatic method of recognition is obviously superior to the earlier method of manual coding and punching of data into computers. Since the set of symbols used in standard musical notation is rather large, a carefully chosen subset of symbols was used in the program.

In this paper, we discuss the problem of automatic recognition of musical notation used in the southern part of India. This music is referred to as Karnatic music to distinguish it from the Hindusthani music of northern India. We are concerned with the recognition of musical scores in Tamil notation where the letters of the Tamil alphabet are used to denote the different notes. Once the notes are recognized they can be translated into western staff notation or any other standard notation. Musicians and researchers who are not familiar with Karnatic music notation will be able to make use of these translations.

Karnatic music is a traditional Indian Music and it is a well developed form of art. Music for singing as well as for instruments is written down in a notation, which is similar to the solfa notation. The octave is divided into seven parts and each part is denoted by a letter of the Tamil alphabet. The actual notes that will be used in a given piece will be denoted by the raga or mode. The notes of the higher octave are distinguished by a dot at the top and those of the lower octave by a dot at the bottom. The problem of recognition of Tamil musical notation reduces primarily to the problem of recognition of letters used in the musical notation.

As in the western notation there are seven main notes in an octave. The names of the notes are abbreviated as Sa, Ri, Ga, Ma, Pa, Dha and Ni. When the first note Sa  corresponds to middle C of the key of  C, the fifth note Pa corresponds to G. Increase in the duration of a note is represented by lengthening of the. vowel in the name. For example, if Sa represented the duration of a crotchet then Sã would represent minim or twice the duration of the note. To denote half the duration, a horizontal line is placed above or below the note, which corresponds to a quaver.

We are concerned here mainly with the question whether it is possible to recognize the pitch and duration of a musical note in Karnatic notation and translate it into staff notation using computer methods. Is it possible to recognize printed music in Karnatic notation? Is it possible to recognize handwritten music if the notes are fairly well formed and written separately? The answer is in the affirmative for both the cases. The problem of recognizing unconstrained cursive handwriting is beyond the scope of this paper.

For recognizing the different characters of modern Tamil alphabet present in Karnatic music notation we use the method of condensed runs2, coded runs3 and their variations4 developed by us. The condensed run method is used in the recognition procedure for printed music. A combination of the other methods is used for the recognition of hand-printed characters.

Each character used in the musical notation is converted into a two level matrix. The value of each entry in the matrix will be '1' or `0' depending upon the colour of the corresponding point of the character. 1 is used to denote the black point and 0 for the white point. The essential features of each character are extracted in the form of strings using the above methods and stored in the memory of a computer. Each character to be recognized is taken up one by one and its features are extracted and compared with the characters already stored. A perfect match yields the direct recognition of the new character.

The major problem in the automatic translation is that of machine recognition of characters. We have shown that it can be achieved for machine as well as hand-printed characters. After recognizing the printed/hand-printed characters, the corresponding notes are printed in the staff notation. The pitch and duration of the musical note can also be printed in the DARMS notation. However in the present experiment we restricted ourselves to getting the translation into staff notation.

A passage is taken from a song in Kalyani raga which makes use of all the seven notes. The letter Sa is treated as equivalent to middle C. It is quite easy to have the correspondence with any other major scale but we have made use of the translation into the C major scale. The translated version could also be displayed on a CRT terminal. Further work is in progress and complete details will be presented in the full paper.

References:

  1. Prerau D. (1971), Computer pattern recognition of printed music, Proceedings of the Fall Joint Computer Conference, pp 153-162.
  2. Gift Siromoney, Chandrasekaran R. and Chandrasekaran M. (1978), Computer recognition of printed Tamil characters, Pattern Recognition, 10, pp 243 -247.
  3. Gift Siromoney, Chandrasekaran R. and Chandrasekaran M. (1978), Computer recognition of an ancient common Indian Script, Proceedings of the Conference on Linguistic Implications of Computer based Information Systems, New Delhi.
  4.  Chandrasekaran M., Chandrasekaran R. and Gift Siromoney(1980), Machine recognition of hand-printed Tamil musical notation, Paper presented at the Annual Conference of the Computer Society of India, Bombay
Go to the top of the page

Home | Biodata | Biography | Photo Gallery | Publications | Tributes