At the outset, every part of the signal is represented many times over using different time scales and angles.

Signal components are then filtered out if they are too small, or not structurally stable. This filter parameter allows the result to scale continuously from a representation that is complete to a representation that captures only the most salient features. We are now working on a script to quantify how much of the signal has been left out at a given filter setting.

Since the image combines contours from many time-scales and angles, aspects of a signal can be represented more than once. (See the discussion on Glottal Pulses and Harmonic stacks.)