This study evaluates the effects of some non-learning blind bandwidth extension (BWE) methods on state-of-the-art automatic speaker verification (ASV) systems. Recently, a non-linear bandwidth extension (N-BWE) method has been proposed as a blind, non-learning, and light-weight BWE approach. Other non-learning BWEs have also been developed in recent years. For ASV evaluations, most data available to train ASV systems is narrowband (NB) telephone speech. Meanwhile, wideband (WB) data have been used to train the state-of-the-art ASV systems, such as i-vector, d-vector, and x-vector. This can cause sampling rate mismatches when all datasets are used. In this paper, we investigate the influence of sampling rate mismatches in the x-vector-based ASV systems and how non-learning BWE methods perform against them. The results showed that the N-BWE method improved the equal error rate (EER) on ASV systems based on the x-vector when the mismatches were present. We researched the relationship between objective measurements and EERs. Consequently, the N-BWE method produced the lowest EERs on both ASV systems and obtained the lower RMS-LSD value and the higher STOI score.

In order to improve the noise robustness of automatic speaker recognition, many techniques on speech/feature enhancement have been explored by using deep neural networks (DNN). In this work, a DNN multi-level enhancement (DNN-ME), which consists of the stages of signal enhancement, cepstrum enhancement and i-vector enhancement, is proposed for text-independent speaker recognition. Given the fact that these enhancement methods are applied in different stages of the speaker recognition pipeline, it is worth exploring the complementary role of these methods, which benefits the understanding of the pros and cons of the enhancements of different stages. In order to use the capabilities of DNN-ME as much as possible, two kinds of methods called Cascaded DNN-ME and joint input of DNNs are studied. Weighted Gaussian mixture models (WGMMs) proposed in our previous work is also applied to further improve the model's performance. Experiments conducted on the Speakers in the Wild (SITW) database have shown that DNN-ME demonstrated significant superiority over the systems with only a single enhancement for noise robust speaker recognition. Compared with the i-vector baseline, the equal error rate (EER) was reduced from 5.75 to 4.01.

Recognizing the different segments of speech belonging to the same speaker is an important speech analysis task in various applications. Recent works have shown that there was an underlying manifold on which speaker utterances live in the model-parameter space. However, most speaker clustering methods work on the Euclidean space, and hence often fail to discover the intrinsic geometrical structure of the data space and fail to use such kind of features. For this problem, we consider to convert the speaker i-vector representation of utterances in the Euclidean space into a network structure constructed based on the local (k) nearest neighbor relationship of these signals. We then propose an efficient community detection model on the speaker content network for clustering signals. The new model is based on the probabilistic community memberships, and is further refined with the idea that: if two connected nodes have a high similarity, their community membership distributions in the model should be made close. This refinement enhances the local invariance assumption, and thus better respects the structure of the underlying manifold than the existing community detection methods. Some experiments are conducted on graphs built from two Chinese speech databases and a NIST 2008 Speaker Recognition Evaluations (SREs). The results provided the insight into the structure of the speakers present in the data and also confirmed the effectiveness of the proposed new method. Our new method yields better performance compared to with the other state-of-the-art clustering algorithms. Metrics for constructing speaker content graph is also discussed.

This paper proposes a speaker-phonetic i-vector modeling method for text-dependent speaker verification with random digit strings, in which enrollment and test utterances are not of the same phrase. The core of the proposed method is making use of digit alignment information in i-vector framework. By utilizing force alignment information, verification scores of the testing trials can be computed in the fixed-phrase situation, in which the compared speech segments between the enrollment and test utterances are of the same phonetic content. Specifically, utterances are segmented into digits, then a unique phonetically-constrained i-vector extractor is applied to obtain speaker and channel variability representation for every digit segment. Probabilistic linear discriminant analysis (PLDA) and s-norm are subsequently used for channel compensation and score normalization respectively. The final score is obtained by combing the digit scores, which are computed by scoring individual digit segments of the test utterance against the corresponding ones of the enrollment. Experimental results on the Part 3 of Robust Speaker Recognition (RSR2015) database demonstrate that the proposed approach significantly outperforms GMM-UBM by 52.3% and 53.5% relative in equal error rate (EER) for male and female respectively.

A text-dependent i-vector extraction scheme and a lexicon-based binary vector (L-vector) representation are proposed to improve the performance of text-dependent speaker verification. I-vector and L-vector are used to represent the utterances for enrollment and test. An improved cosine distance kernel is constructed by combining i-vector and L-vector together and is used to distinguish both speaker identity and lexical (or text) diversity with back-end support vector machine (SVM). Experiments are conducted on RSR 2015 Corpus part 1 and part 2, the results indicate that at most 30% improvement can be obtained compared with traditional i-vector baseline.

Sparse representation has been studied within the field of signal processing as a means of providing a compact form of signal representation. This paper introduces a sparse representation based framework named Sparse Probabilistic Linear Discriminant Analysis in speaker recognition. In this latent variable model, probabilistic linear discriminant analysis is modified to obtain an algorithm for learning overcomplete sparse representations by replacing the Gaussian prior on the factors with Laplace prior that encourages sparseness. For a given speaker signal, the dictionary obtained from this model has good representational power while supporting optimal discrimination of the classes. An expectation-maximization algorithm is derived to train the model with a variational approximation to a range of heavy-tailed distributions whose limit is the Laplace. The variational approximation is also used to compute the likelihood ratio score of all trials of speakers. This approach performed well on the core-extended conditions of the NIST 2010 Speaker Recognition Evaluation, and is competitive compared to the Gaussian Probabilistic Linear Discriminant Analysis, in terms of normalized Decision Cost Function and Equal Error Rate.