Improved quality and intelligibility of mimicking Lombard speech by source-lter and coarticulation model-based synthesis

摘要

In this report, we described the method and pro-posed the extended Lombard model for mimickingLombard speech under varying noise levels. Thecoarticulation, source-lte-based model and Lom-bard effect model of feature values with noise levelshelped to improve similarity and improve intelligi-bility to be equivalent with Lombard speech in anoise level. Further improvement of these models isstill being done to obtain a better result.