Skip to main content

Table 6 Average computation overhead (percent) per frame of the five doctors

From: Integrated exemplar-based template matching and statistical modeling for continuous speech recognition

  All templates (LLR) MLTS (KL) Template compression (KL)
Test frame labeling overhead 43.3 43.3 23.3
Rescoring overhead 26.7 5.4 5.4
Overall computational overhead 70.0 48.7 28.7
  1. Computational overhead per frame using LLR-based all templates and KL-based MLTS and template selection for five doctors in the telehealth task.