Higher order information set based features for text-independent speaker identification

Jeevan Medikonda, Hanmandlu Madasu

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)

Abstract

In this paper Type-2 Information Set (T2IS) features and Hanman Transform (HT) features as Higher Order Information Set (HOIS) based features are proposed for the text independent speaker recognition. The speech signals of different speakers represented by Mel Frequency Cepstral Coefficients (MFCC) are converted into T2IS features and HT features by taking account of the cepstral and temporal possibilistic uncertainties. The features are classified by Improved Hanman Classifier (IHC), Support Vector Machine (SVM) and k-Nearest Neighbours (kNN). The performance of the proposed approaches is tested in terms of speed, computational complexity, memory requirement and accuracy on three datasets namely NIST-2003, VoxForge 2014 speech corpus and VCTK speech corpus and compared with that of the baseline features like MFCC, ∆MFCC, ∆∆MFCC and GFCC under white Gaussian noisy environment at different signal-to-noise ratios. The proposed features have the reduced feature size, computational time, and complexity and also their performance is not degraded under the noisy environment.

Original languageEnglish
Pages (from-to)451-461
Number of pages11
JournalInternational Journal of Speech Technology
Volume21
Issue number3
DOIs
Publication statusPublished - 01-09-2018

All Science Journal Classification (ASJC) codes

  • Software
  • Language and Linguistics
  • Human-Computer Interaction
  • Linguistics and Language
  • Computer Vision and Pattern Recognition

Fingerprint Dive into the research topics of 'Higher order information set based features for text-independent speaker identification'. Together they form a unique fingerprint.

Cite this