Speaker Identification Task Research Articles

METHOD FOR AUTOMATIC SPEAKER RECOGNITION WITH HURST PARAMETER BASED FEATURES AND METHOD FOR SPEAKER CLASSIFICATION BASED ON FRACTIONAL BROWNIAN MOTION CLASSIFIERS

It is proposed a text-independent automatic speaker recognition (ASkR) system which employs a new speech feature and a new classifier. The statistical feature pH is a vector of Hurst parameters obtained by applying a wavelet-based multi-dimensional estimator (M dim wavelets) to the windowed short-time segments of speech. The proposed classifier for the speaker identification and verification tasks is based on the multi-dimensional fBm (fractional Brownian motion) model, denoted by M dim fBm. For a given sequence of input speech features, the speaker model is obtained from the sequence of vectors of H parameters, means and variances of these features.

Colored Noise Based Multicondition Training Technique for Robust Speaker Identification

This letter proposes a colored noise based multicondition training technique for robust speaker identification in unknown noisy environments. The colored noise samples generation is based on filtering a white Gaussian sequence that leads to a power spectral density (PSD) proportional to 1/ <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">f</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">β</sup> , where β ∈ [0, 2]. Gaussian mixture models (GMM) are applied to obtain the speaker models using the noisy speech signals with a single signal-to-noise ratio (SNR). The colored noise based multicondition training is evaluated for the speaker identification task considering the test utterances corrupted with real acoustic noises and different values of SNR. The results show that the proposed technique outperforms the white noise based multicondition and the clean-speech training approaches.

Joint Speech Enhancement and Speaker Identification Using Approximate Bayesian Inference

We present a variational Bayesian algorithm for joint speech enhancement and speaker identification that makes use of speaker dependent speech priors. Our work is built on the intuition that speaker dependent priors would work better than priors that attempt to capture global speech properties. We derive an iterative algorithm that exchanges information between the speech enhancement and speaker identification tasks. With cleaner speech we are able to make better identification decisions and with the speaker dependent priors we are able to improve speech enhancement performance. We present experimental results using the TIMIT data set which confirm the speech enhancement performance of the algorithm by measuring signal-to-noise (SNR) ratio improvement and perceptual quality improvement via the Perceptual Evaluation of Speech Quality (PESQ) score. We also demonstrate the ability of the algorithm to perform voice activity detection (VAD). The experimental results also demonstrate that speaker identification accuracy is improved.

An Auditory-Based Feature Extraction Algorithm for Robust Speaker Identification Under Mismatched Conditions

An auditory-based feature extraction algorithm is presented. We name the new features as cochlear filter cepstral coefficients (CFCCs) which are defined based on a recently developed auditory transform (AT) plus a set of modules to emulate the signal processing functions in the cochlea. The CFCC features are applied to a speaker identification task to address the acoustic mismatch problem between training and testing environments. Usually, the performance of acoustic models trained in clean speech drops significantly when tested in noisy speech. The CFCC features have shown strong robustness in this kind of situation. In our experiments, the CFCC features consistently perform better than the baseline MFCC features under all three mismatched testing conditions-white noise, car noise, and babble noise. For example, in clean conditions, both MFCC and CFCC features perform similarly, over 96%, but when the signal-to-noise ratio (SNR) of the input signal is 6 dB, the accuracy of the MFCC features drops to 41.2%, while the CFCC features still achieve an accuracy of 88.3%. The proposed CFCC features also compare favorably to perceptual linear predictive (PLP) and RASTA-PLP features. The CFCC features consistently perform much better than PLP. Under white noise, the CFCC features are significantly better than RASTA-PLP, while under car and babble noise, the CFCC features provide similar performances to RASTA-PLP.

Robust speaker identification in the presence of car noise

A robust feature set, Teager Energy Operator based Cepstral Coefficients (TEOCC) for speaker identification task is proposed in this paper. Admissible Wavelet Packet (AWP) transform and the Teager Energy Operator (TEO) is used to obtain robust features. The proposed features significantly improve the speaker identification performance (76.25%) compared with the Mel Frequency Cepstral Coefficient (MFCC) features (57%) in the presence of car noise. The performance is evaluated using TIMIT and NOISEX-92 databases. This paper shows that higher-frequency bands also carry more speaker-specific information and the identification rate can be improved without additional processing of the signal to remove noise.

Improving SVM by Modifying Kernel Functions for Speaker Identification Task

Support vector machine (SVM) was the first proposed kernel-based method. It uses a kernel function to transform data from input space into a high-dimensional feature space in which it searches for a separating hyperplane. SVM aims to maximise the generalisation ability that depends on the empirical risk and the complexity of the machine. SVM has been widely adopted in real-world applications including speech recognition. In this paper, an empirical comparison of kernel selection for SVM were used and discussed to achieve performance on text-independent speaker identification using the TIMIT corpus. We were focused on SVM trained using linear, polynomial and radial basis function (RBF) kernels. Results showed that the best performance had been achieved by using polynomial kernel and reported a speaker identification rate equal to 82.47%.

Speaker Recognition Using Auditory Features and Polynomial Classifier

This paper presents a speaker recognition method which makes use of auditory features and polynomial classifier for speaker recognition. Auditory features based on an auditory periphery model extract significant speaker characteristics. Polynomial classifier has been used to accomplish speaker recognition task. Polynomial classifier has several advantages over the conventional classifiers such as computational scalability with the number of speakers, discriminative training allowing it to use out of class data and the statistical interpretation of scoring allowing it to combine with HMM and GMM. This approach achieves substantial performance improvement in a speaker identification task compared with state-of-the-art in a wide range of signal to noise conditions.

On the Importance of Transition Regions for Automatic Speaker Recognition

This letter describes the importance of transition regions, e.g. at phoneme boundaries, for automatic speaker recognition compared with using steady-state regions. Experimental results of automatic speaker identification tasks confirm that transition regions include the most speaker distinctive features. A possible reason for obtaining such results is described in view of articulation, in particular, the degree of freedom of articulators. These results are expected to provide useful information in designing an efficient automatic speaker recognition system.

Acoustic variability and speaker discriminative power of liquids in American English.

In American English, liquid sounds /r/ and /l/ are the most articulatorily variable and complex sounds. In previous work [Zhou et al., J. Acoust. Soc. Am. 123, pp. 4466–4481 (2008)], we found that the acoustic variability of /r/ due to two different tongue shapes (“retroflex” versus “bunched”) is reflected at the fourth and fifth formants. In order to understand how this articulatory diversity contributes to the intra‐ and interspeaker variabilities in speech signals of liquid sounds, an analysis of variance (ANOVA) analysis has been performed to investigate the intra‐ and interspeaker variabilities of liquid sounds across different frequency subbands. The Buckeye database was analyzed in this study, which includes about 30–60 min of broadband conversation speech for each of 40 speakers. The ANOVA analysis showed that liquids have the maximum F‐ratios (interspeaker variance/intraspeaker variance) in the range of about 3–5 kHz, and /l/ has a larger maximum F‐ratio than /r/. In general, liquids have larger F‐ratios than stops and fricatives and have smaller F‐ratios than vowels and nasals. The results of phoneme‐based speaker identification task will be presented to show the speaker discriminative power of liquids along with the results for other phonemes such as vowels and nasals.

Speaker Identification Using Instantaneous Frequencies

This paper presents an experimental evaluation of different features for use in speaker identification. The features are tested using speech data provided by the chains corpus, in a closed-set speaker identification task. The main objective of the paper is to present a novel parametrization of speech that is based on the AM-FM representation of the speech signal and to assess the utility of these features in the context of speaker identification. In order to explore the extent to which different instantaneous frequencies due to the presence of formants and harmonics in the speech signal may predict a speaker's identity, this work evaluates three different decompositions of the speech signal within the same AM-FM framework: a first setup has been used previously for formant tracking, a second setup is designed to enhance familiar resonances below 4000 Hz, and a third setup is designed to approximate the bandwidth scaling of the filters conventionally used in the extraction of Mel-fequency cepstral coefficients (MFCCs). From each of the proposed setups, parameters are extracted and used in a closed-set text-independent speaker identification task. The performance of the new featural representation is compared with results obtained adopting MFCC and RASTA-PLP features in the context of a generic Gaussian mixture model (GMM) classification system. In evaluating the novel features, we look selectively at information for speaker identification contained in the frequency range 0-4000 Hz and 4000-8000 Hz, as the instantaneous frequencies revealed by the AM-FM approach suggest the presence of structures not well known from conventional spectrographic analyses. Accuracy results obtained using the new parametrization perform as well as conventional MFCC parameters within the same reference system, when tested and trained on modally voiced speech which is mismatched in both channel and style. When the testing material is whispered speech, the new parameters provide better results than any of the other features tested, although they remain far from ideal in this limiting case.

Effects of Phoneme Type and Frequency on Distributed Speaker Identification and Verification

In the European Telecommunication Standards Institute (ETSI), Distributed Speech Recognition (DSR) front-end, the distortion added due to feature compression on the front end side increases the variance flooring effect, which in turn increases the identification error rate. The penalty incurred in reducing the bit rate is the degradation in speaker recognition performance. In this paper, we present a nontraditional solution for the previously mentioned problem. To reduce the bit rate, a speech signal is segmented at the client, and the most effective phonemes (determined according to their type and frequency) for speaker recognition are selected and sent to the server. Speaker recognition occurs at the server. Applying this approach to YOHO corpus, we achieved an identification error rate (ER) of 0.05% using an average segment of 20.4% for a testing utterance in a speaker identification task. We also achieved an equal error rate (EER) of 0.42% using an average segment of 15.1% for a testing utterance in a speaker verification task.

Text-independent speaker recognition based on the Hurst parameter and the multidimensional fractional Brownian motion model

In this paper, a text-independent automatic speaker recognition (ASkR) system is proposed-the SR/sub Hurst/-which employs a new speech feature and a new classifier. The statistical feature pH is a vector of Hurst (H) parameters obtained by applying a wavelet-based multidimensional estimator (M/spl I.bar/dim/spl I.bar/wavelets ) to the windowed short-time segments of speech. The proposed classifier for the speaker identification and verification tasks is based on the multidimensional fBm (fractional Brownian motion) model, denoted by M/spl I.bar/dim/spl I.bar/fBm. For a given sequence of input speech features, the speaker model is obtained from the sequence of vectors of H parameters, means, and variances of these features. The performance of the SR/sub Hurst/ was compared to those achieved with the Gaussian mixture models (GMMs), autoregressive vector (AR), and Bhattacharyya distance (dB) classifiers. The speech database-recorded from fixed and cellular phone channels-was uttered by 75 different speakers. The results have shown the superior performance of the M/spl I.bar/dim/spl I.bar/fBm classifier and that the pH feature aggregates new information on the speaker identity. In addition, the proposed classifier employs a much simpler modeling structure as compared to the GMM.

Combining classifier decisions for robust speaker identification

In this work, we combine the decisions of two classifiers as an alternative means of improving the performance of a speaker recognition system in adverse environments. The difference between these classifiers is in their feature-sets. One system is based on the popular mel-frequency cepstral coefficients (MFCC) and the other on the new parametric feature-sets (PFS) algorithm. The feature-vectors both have mel-scale spectral warping and are computed in the cepstral domain but the feature-sets differs in the use of spectral filters and compressions. The performance of the classifier is not much different in recognition rates terms but they are complementary. This shows that there is information that is not captured in the popular mel-frequency cepstral coefficients (MFCC), and the parametric feature-sets (PFS) is able to add further information for improved performance. Several ways of combining these classifiers gives significant improvements in a speaker identification task using a very large telephone degraded NTIMIT database.

Combination of autocorrelation-based features and projection measure technique for speaker identification

This paper presents a robust approach for speaker identification when the speech signal is corrupted by additive noise and channel distortion. Robust features are derived by assuming that the corrupting noise is stationary and the channel effect is fixed during an utterance. A two-step temporal filtering procedure on the autocorrelation sequence is proposed to minimize the effect of additive and convolutional noises. The first step applies a temporal filtering procedure in autocorrelation domain to remove the additive noise, and the second step is to perform the mean subtraction on the filtered autocorrelation sequence in logarithmic spectrum domain to remove the channel effect. No prior knowledge of noise characteristic is necessary. The additive noise can be a colored noise. Then the proposed robust feature is combined with the projection measure technique to gain further improvement in recognition accuracy. Experimental results show that the proposed method can significantly improve the performance of speaker identification task in noisy environment.

A Model-Selection-Based Self-Splitting Gaussian Mixture Learning with Application to Speaker Identification

We propose a self-splitting Gaussian mixture learning (SGML) algorithm for Gaussian mixture modelling. The SGML algorithm is deterministic and is able to find an appropriate number of components of the Gaussian mixture model (GMM) based on a self-splitting validity measure, Bayesian information criterion (BIC). It starts with a single component in the feature space and splits adaptively during the learning process until the most appropriate number of components is found. The SGML algorithm also performs well in learning the GMM with a given component number. In our experiments on clustering of a synthetic data set and the text-independent speaker identification task, we have observed the ability of the SGML for model-based clustering and automatically determining the model complexity of the speaker GMMs for speaker identification.

Speaker recognition using dynamic synapse based neural networks with wavelet processing

We have designed systems utilizing a dynamic synapse neural network (DSNN) to perform speech recognition tasks. The DSNN architecture has demonstrated noise-resistive properties in previous research. The speech signal is first passed through a bank of bandpass filters. The output of each filter then undergoes wavelet transformation to extract speaker-dependent features. Transformed signals are used as input for a DSNN, which is trained using a genetic algorithm training method. We are seeking to explore the capabilities of a biologically based neural network in solving a problem which humans are still better at solving than conventional solutions. The dynamic synapse neural networks were trained to perform several speech recognition tasks, including speaker verification, speaker identification, and word recognition. In the speaker verification task, the objective was to distinguish one speaker from a group of up to 22 imposters based on the utterance of a codeword. In the speaker identification task, the objective was to identify which of 6 speakers has uttered the codeword. In the word recognition task, the objective was to perform speaker-dependent word recognition on a set of 30 words. Neural networks were capable of 90% correct identifications in the presence of white noise (0 dB SNR).

Speaker identification with dynamic synapses

In this paper, we have proposed a new method of presenting a speech signal to the dynamic synapses neural network (DSNN) and optimal selection of input neurons for the speaker identification task. The DSNN developed by Liaw and Berger (1996) was employed to capture the speaker information from the action potentials, which was generated by wavelet filter banks. In order to sufficiently capture the speaker information, the time-delayed temporal patterns of action potentials are used as the inputs of the network. The optimal number of input neurons obtained by minimizing the total output error of the network while identifying a target speaker and rejecting the other unknown speakers. The genetic algorithm was employed to design a supervised learning rule for training of the network. The TI-46 dataset was used in both the training and testing phase. Comparative results with previous methods showed a better performance of the correct classification for the model-based dynamic synapses speaker identification system. [Work supported by DARPA and ONR.]

Identification of twins from pure (single speaker) and hybrid (fused) syllables: an acoustic and perceptual case study.

This study describes a speaker-identification task carried out using two sets of single monosyllables and two sets of hybrid monosyllables prepared from monosyllabic words spoken by a set of monozygotic twins. The task was carried out by 6 listeners who were all very familiar with the twins and by the twins themselves. The identification scores for the pure syllables showed that both twins were identified with accuracy above chance; however, the percentage scores were less than perfect in both cases (71.4% and 72.9%). Analysis of the identification scores for the hybrid syllables showed that listeners had difficulty in attributing the hybrid syllables to a single twin. This was confirmed by the lack of significance in three binomial tests. The pure syllables were analysed acoustically to obtain a set of acoustic parameters for each twin. The results of the perceptual study are discussed with reference to the acoustic parameters. In addition, the implications of the study for issues in speaker identification are considered.

Subband architecture for automatic speaker recognition

We present an original approach for automatic speaker identification especially applicable to environments which cause partial corruption of the frequency spectrum of the signal. The general principle is to split the whole frequency domain into several subbands on which statistical recognizers are independently applied and then recombined to yield a global score and a global recognition decision. The choice of the subband architecture and the recombination strategies are particularly discussed. This techniques had been shown to be robust for speech recognition when a narrow band noise degradation occur. We first objectively verify this robustness for the speaker identification task. We also study which information is really used to recognize speakers. For this, speaker identification experiments on independent subbands are conducted for 630 speakers of TIMIT and NTIMIT databases. The results show that the speaker specific information is not equally distributed among subbands. In particular, the low-frequency subbands (under 600 Hz) and the high-frequency subbands (over 3000 Hz) are more speaker-specific than middle-frequency ones. In addition, experiments on different subband system arechitectures show that the correlations between frequency channels are of prime importance for speaker recognition. Some of these correlations are lost when the frequency domain is divided into subbands. Consequently, we propose a particularly redundant parallel architecture for which most of the correlations are kept. The performances obtained with this new system, using linear recombination strategies, are equivalent to those of a conventional fullband recognizer on clean and telephone speech. Experiments on speech corrupted by unpredictable noise show a better adaptability of this approach in noisy environments, compared to a conventional device, especially when pruning of some recognizers is performed.

IDENTIFICATION OF TWINS FROM PURE (SINGLE SPEAKER) AND HYBRID (FUSED) SYLLABLES: AN ACOUSTIC AND PERCEPTUAL CASE STUDY

This study describes a speaker-identification task carried out using two sets of single monosyllables and two sets of hybrid monosyllables prepared from monosyllabic words spoken by a set of monozygotic twins. The task was carried out by 6 listeners who were all very familiar with the twins and by the twins themselves. The identification scores for the pure syllables showed that both twins were identified with accuracy above chance; however, the percentage scores were less than perfect in both cases (71.4% and 72.9%). Analysis of the identification scores for the hybrid syllables showed that listeners had difficulty in attributing the hybrid syllables to a single twin. This was confirmed by the lack of significance in three binomial tests. The pure syllables were analysed acoustically to obtain a set of acoustic parameters for each twin. The results of the perceptual study are discussed with reference to the acoustic parameters. In addition, the implications of the study for issues in speaker-identification are considered.