Regularized linear prediction of speech
2008 (English)Article in journal (Refereed) Published
All-pole spectral envelope estimates based on linear prediction (LP) for speech signals often exhibit unnaturally sharp peaks, especially for high-pitch speakers. In this paper, regularization is used to penalize rapid changes in the spectral envelope, which improves the spectral envelope estimate. Based on extensive experimental evidence, we conclude that regularized linear prediction outperforms bandwidth-expanded linear prediction. The regularization approach gives lower spectral distortion on average, and fewer outliers, while maintaining a very low computational complexity.
Place, publisher, year, edition, pages
2008. Vol. 16, no 1, 65-73 p.
bandwidth expansion, envelope estimation, linear prediction (LP), regularization, speaker recognition, parameters, systems
IdentifiersURN: urn:nbn:se:kth:diva-17259DOI: 10.1109/tasl.2007.909448ISI: 000251947000007ScopusID: 2-s2.0-64949193677OAI: oai:DiVA.org:kth-17259DiVA: diva2:335302
QC 201005252010-08-052010-08-05Bibliographically approved