Change search
ReferencesLink to record
Permanent link

Direct link
Speech enchancement using hidden Markov models embedded in nonstationary noise
2001 (English)Independent thesis Advanced level (professional degree), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

In mobile communications, the speech signal may be highly degraded by the presence of background noise, e.g. hands free communication in a noisy car. The quality or the intelligibility for human or machine recognizers can be improved with speech enhancement algorithms. Most of the digital signal processing techniques for speech enhancement, as spectral subtraction, require stationary noise to some extent. In many real situations, this constraint is not fulfilled. The speech enhancement method dealt in this thesis is based on stochastic models, Hidden Markov models (HMM´s), modeling the clean speech and noise. The HMM´s accomodates the non-stationarity of the speech and noise with multiples states machines. The information provided by these states machines is used to calculate the filter that is applied to the noisy speech. In the present thesis, this method has been implemented and evaluated with different values for the parameters defining the HMM´s and with different input SNRs.

Place, publisher, year, edition, pages
Keyword [en]
Keyword [sv]
URN: urn:nbn:se:ltu:diva-53496ISRN: LTU-EX--01/238--SELocal ID: a81d5b5d-71a4-4c50-82b1-29aec90a4c47OAI: diva2:1026870
Subject / course
Student thesis, at least 30 credits
Educational program
Electrical Engineering, master's level
Validerat; 20101217 (root)Available from: 2016-10-04 Created: 2016-10-04Bibliographically approved

Open Access in DiVA

No full text

Search outside of DiVA

GoogleGoogle Scholar

Total: 1 hits
ReferencesLink to record
Permanent link

Direct link