Complex Vocal Analysis


    Oct 04 2006 | 2:45 pm
    Hi, im trying to build a patch that will give me a steady real time analysis of vocal sounds independant of pitch, (i.e brightness, timbre) Ive tried using tristan-jehans analyser~ object but it is too general i.e their are too many partials in vocal sounds and the analysis of noise and brightness are always the same. Alsobecause the the object is based around Miller PUckette's fiddle~ the brightness is not completely independant of pitch.
    Can anyone suggest any good methods of anylising vocals that will give me the widest possible range of outputs (which i then intend to map to jitter?)
    cheers
    Will

    • Oct 05 2006 | 12:07 am
      Hi -
      Having tried to do this a few times, you're probably not interested in only the widest range of outputs ... but maybe you are interested in the parameters that humans find it easiest to relate to the sound that is being analysed?
      I'd say getting vibrato rate and extent is probably a good way of looking at vocals. You might be able to do a reasonable job with the hilbert transform applied to the output of a pitch picker like fiddle~ or yin~. I'd like to get this working sometime soon.
      Also - formants in speech are responsible for vowels, and therefore are really easy to relate to singing sound. The lowest two are mostly used for manipulating vowel sound, and the upper two are used for the character of the voice. Has anyone out there got formant analysis working well in real time?
      -- Sam
      On 05/10/2006, at 12:45 AM, WillChadwick wrote:
      > > Hi, im trying to build a patch that will give me a steady real time > analysis of vocal sounds independant of pitch, (i.e brightness, > timbre) > Ive tried using tristan-jehans analyser~ object but it is too > general i.e their are too many partials in vocal sounds and the > analysis of noise and brightness are always the same. Alsobecause > the the object is based around Miller PUckette's fiddle~ the > brightness is not completely independant of pitch. > > Can anyone suggest any good methods of anylising vocals that will > give me the widest possible range of outputs (which i then intend > to map to jitter?) > > cheers > > Will
      --