Automatic CuedSpeech: Overview

The problem of cueing speech is divided into 4 sequential tasks:

1. WHAT? From the time-aligned phonemes, what are the keys?
2. WHEN? From the time-aligned keys, when moving the hand (position and shape)?
3. WHERE? Where should the hand be located, looking at the face?
4. HOW? How the hand should look like?

The "what", "when" and "where" questions does not require any specific feature. The "how" packages require "video" feature of SPPAS to be enabled.