Background noise added to speech can decrease the performance of speech segmentation and enhancement. To solve this problem, new methods have been developed in this thesis. First, a new speech segmentation method (ATF-based SONFIN algorithm) is proposed in fixed noise-level environment. This method contains the multiband analysis and a neural fuzzy network, and it achieves higher recognition rate than the TF-based robust algorithm by 5%. In addition, a new speech segmentation method called RTF-based RSONFIN algorithm is proposed for variable noise-level environment. The RTF-based RSONFIN algorithm contains a recurrent neural fuzzy network. This method contains the multiband analysis and achieve higher recognition rate than the TF-based robust algorithm by 12%.
CITATION STYLE
Lin, C. T., Liu, D. J., Wu, R. C., & Wu, G. D. (2002). Noisy speech segmentation/enhancement with multiband analysis and neural fuzzy networks. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 2275, pp. 301–309). Springer Verlag. https://doi.org/10.1007/3-540-45631-7_40
Mendeley helps you to discover research relevant for your work.