Experimental frameworks

• Example given for the Tulips1 database:

• • Human subjects untrained on lip reading: 89.93%correct
    Sample size: 6 subjects
    
  • Trained lip readers: 95.49% correct
    Sample size: 3 subjects
    
    
    (See [Movellan])

Department of Informatics

Aristotle University of Thessaloniki

• Human speechreading performance:

Previous slide

Next slide

Back to first slide

View graphic version