Optimizing Multi-Taper Features for Deep Speaker Verification

TL;DR

This paper introduces an optimized multi-taper feature extraction method for deep speaker verification, significantly improving robustness and accuracy over traditional static-taper approaches.

Contribution

It proposes jointly optimizing multi-taper estimators with deep neural networks for speaker verification, a novel approach that enhances performance.

Findings

Up to 25.8% EER reduction on SITW corpus

Improved robustness and balanced leakage-variance trade-off

Demonstrated effectiveness of joint optimization in deep ASV

Abstract

Multi-taper estimators provide low-variance power spectrum estimates that can be used in place of the windowed discrete Fourier transform (DFT) to extract speech features such as mel-frequency cepstral coefficients (MFCCs). Even if past work has reported promising automatic speaker verification (ASV) results with Gaussian mixture model-based classifiers, the performance of multi-taper MFCCs with deep ASV systems remains an open question. Instead of a static-taper design, we propose to optimize the multi-taper estimator jointly with a deep neural network trained for ASV tasks. With a maximum improvement on the SITW corpus of 25.8% in terms of equal error rate over the static-taper, our method helps preserve a balanced level of leakage and variance, providing more robustness.