Robust speaker recognition based on filtering in autocorrelation domain and sub-band feature recombination

Quality measures based calibration with duration and noise dependency for speaker recognition

“…This degradation is caused by the lack of information in the short duration condition, and the obstruction of speech information in the occurrence of noise. Many studies have been carried out to evaluate this performance degradation due to short duration (Mandasari et al, 2011;Hasan et al, 2013;Sarkar et al, 2012;Kanagasundaram et al, 2011Kanagasundaram et al, , 2014McLaren et al, 2010) and noise (Mandasari et al, 2012;Togneri and Pullella, 2011;Ming et al, 2007;Kim et al, 2010) in speaker recognition.…”

“…A number of noise robust features such as prosodic (Lei et al, 2012) and auditory (Shao and Wang, 2008) features are used as alternatives to the well-known Mel-frequency cepstrum coefficient (MFCC) features. In addition, modeling and scoring methods have been proposed, e.g., sub-band likelihood scoring (Kim et al, 2010;Kim et al, 2008), universal compensation (UC) model (Ming et al, 2005), and missing http://dx.doi.org/10.1016/j.specom.2015.05.009 0167-6393/Ó 2015 Elsevier B.V. All rights reserved. data technique in universal background model (UBM) (May et al, 2012).…”

Discriminative likelihood score weighting based on acoustic-phonetic classification for speaker identification

“…It is said that some causes of slower progress in speaker identification might be due to the increase in the expected error with growing population size and very high computational cost [3]. Most of the current speaker identification approaches are also based on the same statistical frameworks such as GMM [1,9] or SVM [2,6,10]. Although SVM has shown to be very effective in two-class classification problems such as speaker verification, it may need further algorithmic development in the multi-class tasks including speaker identification [11].…”

The effect of noise on modern automatic speaker recognition systems

“…However, in many applications of speaker recognition, the speech samples provided to the system may suffer from some background noise. In noisy conditions, the performance of speaker recognition system is expected to drop, especially in a low signal-to-noise ratio (SNR) situation [1,2,3].…”

“…In the last decade, some research has studied the behaviour of speaker recognition systems in noisy speech conditions [1], and a number of techniques has been proposed make speaker recognition systems more noise-robust [2,3,4]. However, to the best of our knowledge, there has been no research reported yet on the noiserobustness of the modern i-vector speaker recognition approach that has recently become mainstream in this field.…”