Close talk Speech enhancement in Linear Microphone array for Laptop application
Independent thesis Advanced level (degree of Master (Two Years))Student thesisAlternative title
Close talk Speech enhancement in Linear Microphone array for Laptop application (Swedish)
Now a day’s communication through laptops is drastically increasing in numerous fields. During the communication between person to person through laptops the speech signals is contaminated by the other speech interference signals. In order to enhance the desired speech signal from the noisy environment there are many algorithms are proposed in speech signal processing. This thesis work studies about the suppression of interference signals produced by the surrounding environments for the close talk applications of laptop. In this thesis work uses the three microphones of linearly equi spaced in 3D-co-ordinate system. The speech enhancement algorithms implemented in microphone array were wiener beamforming, Generalized sidelobe canceller using LMS and N-LMS. In order to enhance the desired speech signal with good quality, compares the result of each algorithm using quality metrics like SNR, SNRI and PESQ. The implementation and validation of the algorithms is simulated in Matlab. The quality metrics taken is SNRI and PESQ. In PESQ the output signal is compared with the original clean speech signal and gives the quality measure of the output signal. The SNR tests were conducted for the different input SNR values according to 0dB, 5dB, 10dB, 15dB, 20dB and 25dB. The Simulation result shows that the wiener beamformer effective noise suppression i.e SNRI is 27.9869dB and maintains the speech quality i.e PESQ measurement is 1.459.The effective noise suppression i.e SNRI of the GSC using LMS is 6.0206 dB higher than the wiener beamformer and speech quality is slightly incremental. Comparing the results of GSC using LMS and N-LMS algorithms, The GSC using N-LMS gives the effective noise suppression 3.48dB higher than the GSC using LMS and speech quality i.e PESQ is slightly decreases.
Place, publisher, year, edition, pages
2012. , 54 p.
Speech enhancement, Generalized sidelobe canceller, Wiener beamformer, Normalized least mean square algorithm.
IdentifiersURN: urn:nbn:se:bth-5786Local ID: oai:bth.se:arkivex184B6366932E5DAAC1257A0800765D7DOAI: oai:DiVA.org:bth-5786DiVA: diva2:833188
+46 7230643432015-04-222012-05-242015-06-30Bibliographically approved