Development andEvaluation ofAutomatic -Speaker based- AudioIdentification andSegmentation forBroadcast NewsRecordings Indexation (original) (raw)
Inthis paper, wedescribe anautomatic- speaker based- audio segmentation andidentiJication system for broadcasted newsindexation purposes. WespeciJically focuson speaker identification andaudioscene detection. Speaker identiJication (SI) isbasedonthe state oftheartGaussian mixture models, whereas scene change detection process usestheclassical Bayesian Information Criteria (BIC)andtherecently proposed DISTBICalgorithm. Inthiswork, theeffectiveness of MelFrequency Cepstral coefficients MFCC,Linear Predictive Cepstral Coefficients LPCC,andLogArea Ratio LARcoefficients arecompared forthepurpose of text-independent speaker identification andspeaker basedaudiosegmentation. Both the Fisher Discrimination Ratio- feature analysis andperformance evaluation intermsofcorrect identification rate onthe TIMITdatabase showed that theLPCCoutperforms the other features especially forloworder coefficients. Our experiments onaudio segmentation module showed that theDISTBICsegmentation technique ismorea...
Sign up for access to the world's latest research.
checkGet notified about relevant papers
checkSave papers to use in your research
checkJoin the discussion with peers
checkTrack your impact
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.