Predict hand transition moments with a rule-based system.
It is inspired from the synchronization model in Attina (2005) and the automatic system in (Duchnowski et al. 2000). It is also inspired by (Bratakos et al., 1998): "A critical delay to display the hand cue is +165ms. The max delay is +100ms. A non-significant delay is +33ms.":
Maroula S. Bratakos, Paul Duchnowski and Louis D. Braida Toward the Automatic Generation of Cued Speech Cued Speech Journal VI 1998 p1-37. (c) 1998 National Cued Speech Association, Inc.
The system implements the following rules to estimate [M1;M2]: - Like (Duchnowski et al. 1998), the structure of the key does not matter. - M2 is before A1, like in (Duchnowski et al. 1998) but not that much - The position transition of the first key after a silence (from neutral) is very early before the sound. The 2nd key of a speech segment is also proportionally earlier than the next ones. See predictposgeneric(). - The position transition from the last key of a speech segment to the neutral one (to a long silence), is delayed. See predictposto_neutral(). - They are estimated proportionally to the average duration of observed A1A3 intervals.
