In this paper we propose and evaluate a speaker attribution system using a complete-linkage clustering method. Speaker attribution refers to the annotation of a collection of spoken audio based on speaker identities. This can be achieved using diarization and speaker linking. The main challenge associated with attribution is achieving computational efficiency when dealing with large audio archives. Traditional agglomerative clustering methods with model merging and retraining are not feasible for this purpose. This has motivated the use of linkage clustering methods without retraining. We first propose a diarization system using complete-linkage clustering and show that it outperforms traditional agglomerative and single-linkage clustering based diarization systems with a relative improvement of 40% and 68%, respectively. We then propose a complete-linkage speaker linking system to achieve attribution and demonstrate a 26% relative improvement in attribution error rate (AER) over the single-linkage speaker linking approach.

This paper introduces the Weighted Linear Discriminant Analysis (WLDA) technique, based upon the weighted pairwise Fisher criterion, for the purposes of improving i-vector speaker verification in the presence of high inter-session variability. By taking advantage of the speaker discriminative information that is available in the distances between pairs of speakers clustered in the development i-vector space, the WLDA technique is shown to provide an improvement in speaker verification performance over traditional Linear Discriminant Analysis (LDA) approaches. A similar approach is also taken to extend the recently developed Source Normalised LDA (SNLDA) into Weighted SNLDA (WSNLDA) which, similarly, shows an improvement in speaker verification performance in both matched and mismatched enrolment/verification conditions. Based upon the results presented within this paper using the NIST 2008 Speaker Recognition Evaluation dataset, we believe that both WLDA and WSNLDA are viable as replacement techniques to improve the performance of LDA and SNLDA-based i-vector speaker verification.

In this paper we extend the concept of speaker annotation within a single-recording, or speaker diarization, to a collection wide approach we call speaker attribution.Accordingly, speaker attribution is the task of clustering expectantly homogenous intersession clusters obtained using diarization according to common cross-recording identities.The result of attribution is a collection of spoken audio across multiple recordings attributed to speaker identities.In this paper, an attribution system is proposed using mean-only MAP adaptation of a combined-gender UBM to model clusters from a perfect diarization system, as well as a JFA-based system with session variability compensation.The normalized cross-likelihood ratio is calculated for each pair of clusters to construct an attribution matrix and the complete linkage algorithm is employed to conduct clustering of the inter-session clusters.A matched cluster purity and coverage of 87.1% was obtained on the NIST 2008 SRE corpus.

Robust speaker verification on short utterances remains a key consideration when deploying automatic speaker recognition, as many real world applications often have access to only limited duration speech data. This paper explores how the recent technologies focused around total variability modeling behave when training and testing utterance lengths are reduced. Results are presented which provide a comparison of Joint Factor Analysis (JFA) and i-vector based systems including various compensation techniques; Within-Class Covariance Normalization (WCCN), LDA, Scatter Difference Nuisance Attribute Projection (SDNAP) and Gaussian Probabilistic Linear Discriminant Analysis (GPLDA). Speaker verification performance for utterances with as little as 2 sec of data taken from the NIST Speaker Recognition Evaluations are presented to provide a clearer picture of the current performance characteristics of these techniques in short utterance conditions.

This paper proposes the use of eigenvoice modeling techniques with the Cross Likelihood Ratio (CLR) as a criterion for speaker clustering within a speaker diarization system.The CLR has previously been shown to be a robust decision criterion for speaker clustering using Gaussian Mixture Models.Recently, eigenvoice modeling techniques have become increasingly popular, due to its ability to adequately represent a speaker based on sparse training data, as well as an improved capture of differences in speaker characteristics.This paper hence proposes that it would be beneficial to capitalize on the advantages of eigenvoice modeling in a CLR framework.Results obtained on the 2002 Rich Transcription (RT-02) Evaluation dataset show an improved clustering performance, resulting in a 35.1% relative improvement in the overall Diarization Error Rate (DER) compared to the baseline system.

For several reasons, the Fourier phase domain is less favored than the magnitude domain in signal processing and modeling of speech. To correctly analyze the phase, several factors must be considered and compensated, including the effect of the step size, windowing function and other processing parameters. Building on a review of these factors, this paper investigates a spectral representation based on the Instantaneous Frequency Deviation, but in which the step size between processing frames is used in calculating phase changes, rather than the traditional single sample interval. Reflecting these longer intervals, the term delta-phase spectrum is used to distinguish this from instantaneous derivatives. Experiments show that mel-frequency cepstral coefficients features derived from the delta-phase spectrum (termed Mel-Frequency delta-phase features) can produce broadly similar performance to equivalent magnitude domain features for both voice activity detection and speaker recognition tasks. Further, it is shown that the fusion of the magnitude and phase representations yields performance benefits over either in isolation.

This paper proposes to improve spoken term detection (STD) accuracy by optimizing the figure of merit (FOM). In this paper, the index takes the form of a phonetic posterior-feature matrix. Accuracy is improved by formulating STD as a discriminative training problem and directly optimizing the FOM, through its use as an objective function to train a transformation of the index. The outcome of indexing is then a matrix of enhanced posterior-features that are directly tailored for the STD task. The technique is shown to improve the FOM by up to 13% on held-out data. Additional analysis explores the effect of the technique on phone recognition accuracy, examines the actual values of the learned transform, and demonstrates that using an extended training data set results in further improvement in the FOM.

This study assesses the recently proposed data-driven background dataset refinement technique for speaker verification using alternate SVM feature sets to the GMM supervector features for which it was originally designed. The performance improvements brought about in each trialled SVM configuration demonstrate the versatility of background dataset refinement. This work also extends on the originally proposed technique to exploit support vector coefficients as an impostor suitability metric in the data-driven selection process. Using support vector coefficients improved the performance of the refined datasets in the evaluation of unseen data. Further, attempts are made to exploit the differences in impostor example suitability measures from varying features spaces to provide added robustness.

We propose a new method to characterize a speaker within the Joint Factor Analysis (JFA) framework. Scoring within the JFA framework can be costly and a new method was proposed to produce an accurate score in a fast manner. However, this method is nonsymmetric and performs badly without any score normalization. We propose a new JFA scoring method that is both symmetrical and efficient. In the same way as means of Gaussians can be concatenated to form a supervector, we use several estimates of speaker factors from the eigenvoice space to build a supervector of factors that we call superfactors. We motivate the use of such factors in the current JFA model through comparison with a Tied Factor Analysis model. We show that this method substantially improves the performance of a system that uses only the standard speaker factors to produce scores, and usually outperforms the baseline system. We also show that this method is relatively effective even when score normalization is not an option.

This paper introduces a novel technique to directly optimise the Figure of Merit (FOM) for phonetic spoken term detection. The FOM is a popular measure of STD accuracy, making it an ideal candidate for use as an objective function. A simple linear model is introduced to transform the phone log-posterior probabilities output by a phone classifier to produce enhanced log-posterior features that are more suitable for the STD task. Direct optimisation of the FOM is then performed by training the parameters of this model using a nonlinear gradient descent algorithm. Substantial FOM improvements of 11% relative are achieved on held-out evaluation data, demonstrating the generalisability of the approach.

The QUT-NOISE-TIMIT corpus consists of 600 hours of noisy speech sequences designed to enable a thorough evaluation of voice activity detection (VAD) algorithms across a wide variety of common background noise scenarios. In order to construct the final mixed-speech database, a collection of over 10 hours of background noise was conducted across 10 unique locations covering 5 common noise scenarios, to create the QUT-NOISE corpus. This background noise corpus was then mixed with speech events chosen from the TIMIT clean speech corpus over a wide variety of noise lengths, signal-to-noise ratios (SNRs) and active speech proportions to form the mixed-speech QUT-NOISE-TIMIT corpus. The evaluation of five baseline VAD systems on the QUT-NOISE-TIMIT corpus is conducted to validate the data and show that the variety of noise available will allow for better evaluation of VAD systems than existing approaches in the literature.

This paper presents a method of voice activity detection (VAD) for high noise scenarios, using a noise robust voiced speech detection feature. The developed method is based on the fusion of two systems. The first system utilises the maximum peak of the normalised time-domain autocorrelation function (MaxPeak). The second zone system uses a novel combination of cross-correlation and zero-crossing rate of the normalised autocorrelation to approximate a measure of signal pitch and periodicity (CrossCorr) that is hypothesised to be noise robust. The score outputs by the two systems are then merged using weighted sum fusion to create the proposed autocorrelation zero-crossing rate (AZR) VAD. Accuracy of AZR was compared to state of the art and standardised VAD methods and was shown to outperform the best performing system with an average relative improvement of 24.8% in half-total error rate (HTER) on the QUT-NOISE-TIMIT database created using real recordings from high-noise environments.

This paper proposes the use of the Bayes Factor as a distance metric for speaker segmentation within a speaker diarization system. The proposed approach uses a pair of constant sized, sliding windows to compute the value of the Bayes Factor between the adjacent windows over the entire audio. Results obtained on the 2002 Rich Transcription Evaluation dataset show an improved segmentation performance compared to previous approaches reported in literature using the Generalized Likelihood Ratio. When applied in a speaker diarization system, this approach results in a 5.1% relative improvement in the overall Diarization Error Rate compared to the baseline.

The recently proposed data-driven background dataset refinement technique provides a means of selecting an informative background for support vector machine (SVM)-based speaker verification systems. This paper investigates the characteristics of the impostor examples in such highly informative background datasets. Data-driven dataset refinement individually evaluates the suitability of candidate impostor examples for the SVM background prior to selecting the highest-ranking examples as arefinedbackground dataset. Further, the characteristics of the refined dataset were analyzed to investigate the desired traits of an informative SVM background. The most informative examples of the refined dataset were found to consist of large amounts of active speech and distinctive language characteristics. The data-driven refinement technique was shown to filter the set of candidate impostor examples to produce a more disperse representation of the impostor population in the SVM kernel space, thereby reducing the number of redundant and less-informative examples in the background dataset. Furthermore, data-driven refinement was shown to provide performance gains when applied to the difficult task of refining a small candidate dataset that was mismatched to the evaluation conditions.

Proposed is an approach to estimating confidence measures on the verification score produced by a Gaussian mixture model (GMM)-based automatic speaker verification system with applications to drastically reducing the typical data requirements for producing a confident verification decision. The confidence measures are based on estimating the distribution of the observed frame scores. The confidence estimation procedure is also extended to produce robust results with very limited and highly correlated frame scores as well as in the presence of score normalization. The proposed Early Verification Decision method utilizes the developed confidence measures in a sequential hypothesis testing framework, demonstrating that as little as 2–10 s of speech on average was able to produce verification results approaching that of using an average of over 100 s of speech on the 2005 NIST SRE protocol.

This paper compares two of the leading techniques for session variability compensation in the context of support vector machine (SVM) speaker verification using Gaussian mixture model (GMM) mean supervectors: joint factor analysis (JFA) modeling and nuisance attribute projection (NAP). Motivation for this comparison comes from the distinctly different domains in which these techniques are employed-the probabilistic GMM domain versus the discriminative SVM kernel. A theoretical analysis is given comparing the JFA and NAP approaches to variability compensation. The role of speaker factors in the factor analysis model is also contrasted against the scatter difference NAP objective of retaining speaker information in the SVM kernel space. These methods for retaining speaker variation are found to provide improved verification performance over the removal of channel effects alone. Overall, experimental results on the NIST 2006 and 2008 SRE corpora demonstrate the effectiveness of both JFA and NAP techniques for reducing the effects of variability. However, the overheads associated with the implementation of JFA may make NAP a more attractive technique due to its simple yet effective approach to variability compensation.

The problem of background dataset selection in SVM-based speaker verification is addressed through the proposal of a new data-driven selection technique. Based on support vector selection, the proposed approach introduces a method to individually assess the suitability of each candidate impostor example for use in the background dataset. The technique can then produce a refined background dataset by selecting only the most informative impostor examples. Improvements of 13% in min. DCF and 10% in EER were found on the SRE 2006 development corpus when using the proposed method over the best heuristically chosen set. The technique was also shown to generalise to the unseen NIST 2008 SRE corpus.

This work aims to take advantage of recent developments in joint factor analysis (JFA) in the context of a phonetically conditioned GMM speaker verification system. Previous work has shown performance advantages through phonetic conditioning, but this has not been shown to date with the JFA framework. Our focus is particularly on strategies for combining the phone-conditioned systems. We show that the classic fusion of the scores is suboptimal when using multiple GMM systems. We investigate several combination strategies in the model space, and demonstrate improvement over score-level combination as well as over a non-phonetic baseline system. This work was conducted during the 2008 CLSP Workshop at Johns Hopkins University.

While spoken term detection (STD) systems based on word indices provide good accuracy, there are several practical applications where it is infeasible or too costly to employ an LVCSR engine. An STD system is presented, which is designed to incorporate a fast phonetic decoding front-end and be robust to decoding errors whilst still allowing for rapid search speeds. This goal is achieved through monophone open-loop decoding coupled with fast hierarchical phone lattice search. Results demonstrate that an STD system that is designed with the constraint of a fast and simple phonetic decoding front-end requires a compromise to be made between search speed and search accuracy.

The problem of impostor dataset selection for GMM-based speaker verification is addressed through the recently proposed data-driven background dataset refinement technique. The SVM-based refinement technique selects from a candidate impostor dataset those examples that are most frequently selected as support vectors when training a set of SVMs on a development corpus. This study demonstrates the versatility of dataset refinement in the task of selecting suitable impostor datasets for use in GMM-based speaker verification. The use of refined Z- and T-norm datasets provided performance gains of 15% in EER in the NIST 2006 SRE over the use of heuristically selected datasets. The refined datasets were shown to generalise well to the unseen data of the NIST 2008 SRE.

This work presents an extended Joint Factor Analysis model including explicit modelling of unwanted within-session variability. The goals of the proposed extended JFA model are to improve verification performance with short utterances by compensating for the effects of limited or imbalanced phonetic coverage, and to produce a flexible JFA model that is effective over a wide range of utterance lengths without adjusting model parameters such as retraining session subspaces. Experimental results on the 2006 NIST SRE corpus demonstrate the flexibility of the proposed model by providing competitive results over a wide range of utterance lengths without retraining and also yielding modest improvements in a number of conditions over current state-of-the-art.

International audience

Training the speaker and session subspaces is an integral problem in developing a joint factor analysis GMM speaker verification system. This work investigates and compares several alternative procedures for this task with a particular focus on training and testing with short utterances. Experiments show that better performance can be obtained when an independent rather than simultaneous optimisation of the two core variability subspaces is used. It is additionally shown that for verification trials on short utterances it is important for the session subspace to be trained with matched length utterances. Conversely, the speaker transform should always be trained with as much data as possible. Index Terms: speaker verification, factor analysis, probabilistic PCA.

Drastic reductions in the typical data requirements for produc-ing confident decisions in an automatic speaker verification sys-tem are demonstrated through the application of a novel ap-proach of score confidence interval estimation. The confidence estimation procedure is also extended to produce robust re-sults with very limited and highly correlated frame scores. The early verification decision method evaluated on the 2005 NIST SRE protocol demonstrates that an average of 2–10 seconds of speech is sufficient to produce verification results approaching those achieved previously using an average of over 100 seconds of speech. Index Terms: automatic speaker verification, confidence mea-sures, verification decision confidence.

This paper compares two of the leading techniques for session variability compensation in the context of GMM mean supervector SVM classifiers for speaker recognition: inter-session variability modelling and nuisance attribute projection.The former is incorporated in the GMM model training while the latter is employed as a modified SVM kernel.Results on both the NIST 2005 and 2006 corpora demonstrate the effectiveness of both techniques for reducing the effects of session variation.Further, system-and score-level fusion experiments show that the combination of the two methods provides improved performance.

This paper details the submission from the Speech and Audio Research Lab of Queensland University of Technology (QUT) to the inaugural 2006 NIST Spoken Term Detection Evaluation. The task involved accurately locating the occurrences of a specified list of English terms in a given corpus of broadcast news and conversational telephone speech. The QUT system uses phonetic decoding and Dynamic Match Lattice Spotting to rapidly locate search terms, combined with a neural network-based verification stage. The use of phonetic search means the system is open vocabulary and performs usefully (Actual Term-Weighted Value of 0.23) whilst avoiding the cost of a large vocabulary speech recognition engine.

Presented is an approach to modelling session variability for GMM-based text-independent speaker verification incorporating a constrained session variability component in both the training and testing procedures. The proposed technique reduces the data labelling requirements and removes discrete categorisation needed by previous techniques and provides superior performance. Experiments on Mixer conversational telephony data show improvements of as much as 46% in equal error rate over a baseline system. In this paper the algorithm used for the enrollment procedure is described in detail. Results are also presented investigating the response of the technique to short test utterances and varying session subspace dimension

This paper examines the usefulness of a multilingual broad syllable-based framework for text-independent speaker verification.Syllabic segmentation is used in order to obtain a convenient unit for constrained and more detailed model generation.Gaussian mixture models are chosen as a suitable modelling paradigm for initial testing of the framework.Promising results are presented for the NIST 2003 speaker recognition evaluation corpus.The syllable-based modelling technique is shown to outperform a state-of-the-art baseline GMM system.A simple selective reduction of the syllable set is also shown to give further improvement in performance.Overall, the syllable based framework presents itself as valid alternative to text-constrained speaker verification systems, with the advantage of being multilingual.The framework allows for future testing of alternative modelling paradigms, feature sets and qualitative analysis.

Handset and channel mismatch degrades the performance of automatic speaker recognition systems significantly. This paper enhances the feature mapping technique by proposing an iterative clustering approach to context model generation which offers an improvement in the performance of feature mapping trained on labelled data and offers the potential to train feature mapping in the absence of correctly labelled background data. The performance of the clustered feature mapping models is demonstrated on an expanded version of the NIST 2003 Extended Data Task (EDT) protocol.

Presented is an approach to modelling session variability for GMM-based text-independent speaker verification incorporating a constrained session variability component in both the training and testing procedures. The proposed technique reduces the data labelling requirements and removes discrete categorisation needed by techniques such as feature mapping and H-Norm, while providing superior performance. Experiments on Switchboard-II conversational telephony data show improvements of as much as 48% in detection cost with a single training utterance and 68% with multiple training utterances over a baseline system.