The paper expounds the language phenomena of English reduction, including the analysis of reduction pattern as to the different aspects of word structure, sentence structure, and figures of speech.
Experiments show a 55.0% reduction in the frame error rate and a 98.5% reduction in sentence breaking error rate with the multimodal VAD as compared to the frame energy-based audio VAD.