A perceptually-motivated low-complexity instantaneous linearchannel normalization technique applied to speaker verification

Abstract

This paper proposes a new set of speech features called Locally-Normalized Cepstral Coefficients (LNCC) that are based onSeneff’s Generalized Synchrony Detector (GSD). First, an analysis of the GSD frequency response is provided to show that itgenerates spurious peaks at harmonics of the detected frequency. Then, the GSD frequency response is modeled as a quotient of twofilters centered at the detected frequency. The numerator is a triangular band pass filter centered around a particular frequency similarto the ordinary Mel filters. The denominator term is a filter that responds maximally to frequency components on either side of thenumerator filter. As a result, a local normalization is performed without the spurious peaks of the original GSD. Speaker verificationresults demonstrate that the proposed LNCC features are of low computational complexity and far more effectively compensate forspectral tilt than ordinary MFCC coefficients. LNCC features do not require the computation and storage of a moving average of thefeature values, and they provide relative reductions in Equal Error Rate (EER) as high as 47.7%, 34.0% or 25.8% when comparedwith MFCC, MFCC + CMN, or MFCC + RASTA in one case of variable spectral tilt, respectively.