Data augmentation versus noise compensation for x-vector speaker recognition systems in noisy environments

Abstract

The explosion of available speech data and new speaker modeling methods based on deep neural networks (DNN) have given the ability to develop more robust speaker recognition systems. Among DNN speaker modelling techniques, x-vector system has shown a degree of robustness in noisy environments. Previous studies suggest that by increasing the number of speakers in the training data and using data augmentation more robust speaker recognition systems are achievable in noisy environments. In this work, we want to know if explicit noise compensation techniques continue to be effective despite the general noise robustness of these systems. For this study, we will use two different x-vector networks: the first one is trained on Voxceleb1 (Protocol1), and the second one is trained on Voxceleb1+Voxveleb2 (Protocol2).We propose to add a denoising x-vector subsystem before scoring. Experimental results show that, the x-vector system used in Protocol2 is more robust than the other one used Protocol1. Despite this observation we will show that explicit noise compensation gives almost the same EER relative gain in both protocols. For example, in the Protocol2 we have 21% to 66% improvement of EER with denoising techniques.

Compensate multiple distortions for speaker recognition systems

“…In [4] the importance of using denoising techniques alongside data augmentation was explored to make the xvector system robust against additive noise. The 66% relative improvement of EER in the x-vector domain shows that noise compensation in speaker modeling level is very effective for speaker recognition systems.…”

“…Using a huge amount of data makes the DNN based systems robust against noise and reverberation but in severe conditions the performance can degrade drastically [3]. Therefore, explicit denoising and dereverberation can make these systems more robust [4,5].…”

“…In the previous work [4,7] we proposed several Denoising Autoencoders (DAE) to reduce the effect of additive noise in x-vector space. The proposed DAEs estimate the clean x-vector from the noisy version.…”

“…Indeed, the proposed DAE tries to learn the relation between the x-vector affected by a given distortion and its clean version. Thus, it is a direct learning of the distortion effect, which would make the denoising system more efficient than multi-condition training [4]. The DAE uses more specific information about the distortion than what would be used in a data augmentation approach.…”

ERIL: An Algorithm for Emotion Recognition From Indian Languages Using Machine Learning

“…Deep Learning offers i-vector and x-vector features [15] [16]. These features are a fusion of MFCC, DWT and MFCC, Gammatone Frequency Cepstral Coefficients (GFCC) respectively.…”