Noise Invariant Frame Selection: A Simple Method to Address the Background Noise Problem for Text-independent Speaker Verification

Abstract

The performance of speaker-related systems usually degrades heavily in practical applications largely due to the presence of background noise. To improve the robustness of such systems in unknown noisy environments, this paper proposes a simple pre-processing method called Noise Invariant Frame Selection (NIFS). Based on several noisy constraints, it selects noise invariant frames from utterances to represent speakers. Experiments conducted on the TIMIT database showed that the NIFS can significantly improve the performance of Vector Quantization (VQ), Gaussian Mixture Model-Universal Background Model (GMM-UBM) and i-vector-based speaker verification systems in different unknown noisy environments with different SNRs, in comparison to their baselines. Meanwhile, the proposed NIFS-based speaker verification systems achieves similar performance when we change the constraints (hyperparameters) or features, which indicates that it is robust and easy to reproduce. Since NIFS is designed as a general algorithm, it could be further applied to other similar tasks.

Noise Robust Speaker Recognition Based on Adaptive Frame Weighting in GMM for i-Vector Extraction

“…In Section III, weighted GMM for i-vector extraction has been presented. In this section, we describe how to calculate different weights for different frames as shown in Fig.2 Inspired by [24], in our proposed algorithm, different types of noises are added to the original testing speech to explore the noise-robustness of different frames. In order to make straightforward comparison w.r.t.…”

“…In order to make straightforward comparison w.r.t. [24], in the following experiments, three types of noises, i.e. white, babble, and pink, are chosen the same as in [24].…”

“…Recently, an algorithm, Noise Invariant Frame Selection (NIFS), was proposed in [24], where the input testing utterance was artificially-noised to help choose noise invariant frames. Experiments showed the effectiveness of this algorithm on speaker recognition in noisy environments, except under low signal-to-noise ratios (SNRs).…”

“…1) An improved Baum-Welch algorithm to train GMM with data weighting, which outperformed a similar approach presented in [27]. 2) A straightforward approach to evaluate the noise robustness of speech frames without additional manual efforts on choosing a threshold in [24]. The remaining part of this paper is organized as follows.…”

Two-stage iterative Procrustes match algorithm and its application for VQ-based verification

“…To solve this problem, speech enhancement 4-6 and model adaptation approaches 7,8 have been widely used. Since the same noise may pose different distortion on different frames, by adopting several noisy constraints, Song et al 9 designed a simple framework that can select noise invariant frames from original audio signals. The experiment results show that this framework can enhance the speaker verification performance for different speaker models under different noisy conditions.…”