Frequency scaling is achieved by multiplicatively modifying the peak frequencies, and frequency shifting by additively modifying the peak frequencies. Frequency scaling occurs before frequency shifting, so the synthesized frequency is
The number of samples generated is unaffected by changes in the component frequencies. Thus, independent control of the time and frequency scale of a synthesized signal can be easily exercised in the domain of the Lemur model. Since the phase curve is computed at the time of synthesis, even radical modifications produce no phase discontinuities or wide frequency excursions, which produce audible frequency artifacts in other sinusoidal methods (Serra 1989).
Finally, track magnitudes may also be controlled during Lemur synthesis. Magnitude scaling is specified in decibels, where a 6dB change in magnitude produces a halving (-6dB) or doubling (+6dB) of the sample values.
Frequency scaling, frequency shifting, and magnitude scaling may optionally be applied to only those tracks in a Lemur analysis bearing a specified label. Time-varying control of scaling and shifting is achieved using control files.
The Time Scale, Frequency Scale, Frequency Shift, and Magnitude Scale for Lemur synthesis may be specified in the Modification Parameters panel, accessed from Lemur's Parameters menu.
A small jazz combo playing up-tempo, synthesized in Lemur:
A small jazz combo, laying back a bit:
What happened to the vibraphone attacks?
Why's the snare so mushy?
Noisy soinds like those are not well modeled by sinusoids. For an explanation of the problem and a brief description of possible solutions, including the solution we are working on, have a look at our ICMC paper from 1995. This is a major concern of Kelly's dissertation work (in progress), stay tuned...