Indian Journal of Science and Technology
DOI: 10.17485/ijst/2016/v9i38/102972
Year: 2016, Volume: 9, Issue: 38, Pages: 1-8
Original Article
Shipra1 * and Mahesh Chandra2
1 Electronics and Communication Engineering Department, BIT Mesra, Near Patna Airport, Patna - 800014, Bihar, India; [email protected]
2 Electronics and Communication Engineering Department, BIT Mesra, Ranchi - 835215, Jharkhand, India; [email protected]
*Author for correspondence
Shipra
Electronics and Communication Engineering Department
Email:[email protected]
This paper present a novel hybridized QCN-PNCC features. These features are obtained by processing Power Normalized Cepstral Coefficients (PNCC) with Quantile based Dynamic Cepstral Normalization Technique (QCN). The robustness of the QCN-PNCC features is compared with PNCC features for the task of Hindi Vowel classification with HMM classifier for Context-Dependent and Context- Independent cases in clean as well as in noisy environment. It is observed that the recognition accuracy of QCN-PNCC features with Hidden Markov Model (HMM) as classifier exhibit an improvement of approximately 8% as compared to PNCC features for Hindi vowel classification task.
Keywords: Power normalized Cepstral Coefficient (PNCC), QCN, QCN-PNCC, Speech Recognition
Subscribe now for latest articles and news.
