DocumentCode :
3668044
Title :
Analysis of 0dB and 10dB babble noise on stuttered speech
Author :
Salma Jabeen;K.M. RaviKumar
Author_Institution :
Dept. of Computer Science and Engg., Ghousia College of Engineering, Ramanagara, Affliated to VTU, India
fYear :
2015
Firstpage :
1
Lastpage :
5
Abstract :
This paper describes the effect of 0dB and 10dB babble noise on stuttered speech. 100 samples are collected from the subjects (stutterer), among which 80 samples are used to prepare database dictionary and 20 samples for testing. The samples are segmented and pass through Mel Frequency Cepstral Coefficient (MFCC) feature extraction and statistical estimation to prepare dictionary, then testing samples compared with the database. The results obtained are 100%. When the same testing samples are added with the babble noise of 0dB and 10dB, the result degrades; the degraded results are enhanced by using sparse matrix enhancement technique. The results obtained after enhancement are 60-95% depending on the samples used for analysis.
Keywords :
"Speech","Noise","Mel frequency cepstral coefficient","Speech recognition","Testing","Noise measurement","Databases"
Publisher :
ieee
Conference_Titel :
Soft-Computing and Networks Security (ICSNS), 2015 International Conference on
Print_ISBN :
978-1-4799-1752-5
Type :
conf
DOI :
10.1109/ICSNS.2015.7292422
Filename :
7292422
Link To Document :
بازگشت