György Szaszák, Klára Vicsi
Using prosody for the improvement of automatic speech recognition
This paper describes sentence, phrase and word boundary detection based on prosodic features, implemented in a HMMbased prosodic segmentation tool. Integrated into a speech recognizer, an N-best rescoring is performed based on the output of the prosodic segmenter, which determines the prosodic structure of the utterance. In an ultrasonography task, we obtained 3,82% speech recognition error reduction using a simplified bi-gram language model.