A novel approach to the development of an efficient speech endpoint detector (EPD) has been proposed to be applied to quasi-continuous and continuous speech processing. The goal is the improvement of multiple-threshold traditional EPDs to fit a larger applications range. Fuzzy logic processing has been applied to implement the decision logic needed to evaluate uncertain data related to transition points among contiguous speech units. Results of the fuzzy logic-based EPD has been compared to the multiple threshold-based EPDs. Extensions towards phonetic unit segmentation and a practical application in virtual agents animation have been discussed.
Improving speech endpoint detection using fuzzy logic-based methodologies / M. Malcangi - In: Proceedings of the Thirtheenth Turkish Symposium on Artificial Intelligence and Neural NetworksIzmir : Dokuz Eylul Universitesi, 2004. - ISBN 9754412138. - pp. 139-146 (( Intervento presentato al 13th. convegno Turkish Symposium on Artificial Intelligence and Neural Networks tenutosi a Foca - Izmir - Turkey nel 2004.
Improving speech endpoint detection using fuzzy logic-based methodologies
M. MalcangiPrimo
2004
Abstract
A novel approach to the development of an efficient speech endpoint detector (EPD) has been proposed to be applied to quasi-continuous and continuous speech processing. The goal is the improvement of multiple-threshold traditional EPDs to fit a larger applications range. Fuzzy logic processing has been applied to implement the decision logic needed to evaluate uncertain data related to transition points among contiguous speech units. Results of the fuzzy logic-based EPD has been compared to the multiple threshold-based EPDs. Extensions towards phonetic unit segmentation and a practical application in virtual agents animation have been discussed.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.