Re: On verification of chandassu
Nasy Sankagiri (narayans@dunx1.ocs.drexel.edu)
Thu, 23 Nov 1995 16:04:48 -0500
Sri Pillalamarri wrote:
>I liked your post on rAga identification. After I sent my mail where I
>compared identifying the chandas to that of identifying a rAga, I realized
>that identifying a rAga, given the ArOhaNa and avarOhaNa notes is a trivial
>task. But a given song in a given rAga (I am treading into unfamiliar
>territory here, please correct me when I make a mis-statement - this is a
>general plea to the musically-talented, from one who is musically-challenged)
>doesn't have the notes repating in the same manner. I believe the notes can
>go back and forth and partial sequences of them can repeat and so on. In such
>a case, the problem becomes more complicated.
>
>But even more complicating would be the problem of "identifying" or
>"recognizing" the sequence of notes that occur in a given song, given
>an audio signal, from some medium. How accurately would the notes be
>identified?
>
I think you had put your finger right on the center, sir. Given an aarOhaNa
and avarOhaNa, identifying the raagam is no big deal. We don't even need to
go into identifying raagam from song, just trying to identify the
swara-pattern from, say, a chiTTa swaram (the series of notes composed by
the composer, mostly prevalent in compositions like varNams) will be
complicated enough. I think it is no mean task to separate the
aarOhaNa-avarOhaNa from a sequence of swaras. Moreover, there are several
raagams which have the same notes. And then, there is the problem of
differentiating the various multiple notes, which 'ri', which 'ga', so on.
All these problems pale in comparison to the problem of SRti. All the above
tasks can be performed by a machine if the music is produced, say, on a
keyboard. However, if the music is played on an instrument like veeNa,
where the SRti is changeable, the same raagam can be produced with
different base tones-how can the pattern-recognizing machine identify the
notes in this case, assuming that it identifies the notes based on
frequency? I read in some 'intro to Karnatic music' book that it is the
ratio of frequencies which defines the places of the notes, and not the
actual frequency itself; it is all relative, in the case of our music, to
the notes 'sa' and 'pa'.
Apart from all this, I am a firm believer that swarams, though they
constitute a raagam, do not make the raagam; it is the emphasis, the
snagathis, sanchaaras, and gamakams which produce what we call raagam. So
if we want to identify raagams, say, by connecting the signal from a CD to
a pattern-matching machine, I guess, the best bet would be to store
excerpts of all the known raagams as rendered by great vidwaans, and see if
any bit in the test piece matches anything in the standard pieces:-)
Sorry, if I took off completely at tangent on this. I wanted to reply to
Mr.Parigi's post itself, but was busy at the moment. Just had to get it
out. Thanks for listening.
Regards
Nasy