[Clam-devel] Re: A YES/NO question about the capability of the synthesizer & validity of the score

Han, Yushen yushen.han at gmail.com
Thu Jul 3 09:41:46 PDT 2008

Hi, Greg

Thank you for your information about how you generate the score.

I had a pitch tracker too which works fairly well for any pitched
sound, but its performance in transition is not tested.

I agree that a score generator that "has access to the original signal
and can make a much better guess as to whether a frame is a transient
than the synthesizer can". For the sake of testing, I will make one

However, I was thinking the program would take input from  MIDI
keyboard + breath controller. (Yours will take the input from ebow,
which will not explicitly say "note of a certain pitch". Is that
correct?). In case that I have the definite pitch ( and note on/note
off ), probably I do not need to worry about note detection. If our
goal is to mimic the oboe sound, I think it will be good to have
well-defined note pitch.
(In the light of Pau's reply yesterday that my program should generate
sound for whatever input, it seems that your way to track the note
information is more universal and should be used.)

I also thought about holding 2 (or more) frames in synthesis to make a
decision on whether it is a new note. A delay of 2 frame (0.011s)
would not be noticed by human but neither it is enough for us to be
sure about whether a new note is coming. You did not hold any frames
in synthesis, did you?

Best regards,
Han, Yushen

On Thu, Jul 3, 2008 at 6:08 AM, Greg Kellum <greg.kellum at gmail.com> wrote:
> Hi Yushen,
> It sounds like you're having problems with transients...  When I
> generated that score for the violin sonata, I used a pitch tracker I
> had been working on earlier to generate the initial score, but then, I
> went back over it and made corrections to the score by hand.
> Sometimes when there were transitions between notes, there simply
> isn't any unambiguous pitch value, and if you force a pitch tracker to
> give you a value for those frames, it will make a mistake, right?
> I later added transient detection to the pitch tracker so if there was
> a big change in the pitch value and the aperiodicity of the signal
> jumped as well, then the new pitch value would be marked as a
> transient and suppressed.  You could do something like this in the
> synthesizer instead.  But the score generator has access to the
> original signal and can make a much better guess as to whether a frame
> is a transient than the synthesizer can.
> Best,
> Greg
> On Thu, Jul 3, 2008 at 12:23 AM, Han, Yushen <yushen.han at gmail.com> wrote:
>> Pau,
>> So it is YES - my synthesizer should be able to play ANYTHING
>> indicated from the score.
>> A difficulty is that I did not know how something that is arbitrarily
>> given in the score file should sound like.
>> The concept of a "note" which should last at least for 50ms, may not
>> apply anymore.
>> But I can still make it work by SMS transformation. Let's hear how it
>> will sound...
>> Han, Yushen
>> On Wed, Jul 2, 2008 at 6:17 PM, Pau Arumí Albó <parumi at iua.upf.edu> wrote:
>>> El dc 02 de 07 de 2008 a les 18:12 -0400, en/na Han, Yushen wrote:
>>>> Here I have a fundamental question about the purpose of this synthesizer:
>>>> Should the synthesizer be able to produce whatever the score indicates,
>>>> even if it is something a real instrument / human musician would not
>>>> be able to do?
>>> sure

More information about the clam-devel mailing list