Speaker Identification Method Research Articles

Soft‐decision a priori knowledge interpolation for robust telephone speaker identification

Handsets which are not seen in the training phase (a.k.a unseen handsets) are main sources of performance degradation for speaker identification (SID) applications in telecommunication environments. To alleviate the problem, a soft‐decision a priori knowledge interpolation (SD‐AKI) method of handset characteristic estimation for handset mismatch‐compensated SID is proposed in this paper. The idea of the SD‐AKI method is to first collect a set of characteristics of seen handsets in the training phase, and to then estimate the characteristic of the unknown testing handset by interpolating the set of seen handset characteristics in the test phase. The estimated handset characteristic is then used to compensate for handset mismatch for robust SID. The SD‐AKI method can be realized in both feature and model spaces. Experimental results on the handset TIMIT (HTIMIT) database showed that both the proposed feature‐ and model‐space SD‐AKI schemes were more robust than the blind cepstral mean subtraction (CMS), feature warping (FW) methods and their hard‐decision counterpart (HD‐AKI) for both cases of all‐handset and unseen‐handset SID tests. It is therefore a promising robust SID method.

Semi-supervised speaker identification under covariate shift

In this paper, we propose a novel semi-supervised speaker identification method that can alleviate the influence of non-stationarity such as session dependent variation, the recording environment change, and physical conditions/emotions. We assume that the voice quality variants follow the covariate shift model, where only the voice feature distribution changes in the training and test phases. Our method consists of weighted versions of kernel logistic regression and cross validation and is theoretically shown to have the capability of alleviating the influence of covariate shift. We experimentally show through text-independent/dependent speaker identification simulations that the proposed method is promising in dealing with variations in voice quality.

Advanced time-frequency displays applied to forensic speaker identification.

Forensic phoneticians have long utilized conventional spectrograms as the basis of the the auditory-acoustic method of speaker identification. Recent investigations [S. A. Fulop and K. Fitz, “Using the reassigned spectrogram to obtain a voiceprint,” J. Acoust. Soc. Am. 119, 3337 (2006); S. A. Fulop and S. Disner, “The reassigned spectrogram as a tool for voice identification,” Int. Congress of Phonetic Sciences XVI, 1853–1856 (2007)] have suggested that a newer technology, i.e., the reassigned spectrogram, may be at least as valuable in the service of speaker identification. The present paper describes methods for forensic speaker identification using enhanced time-frequency displays including reassigned spectrograms and Zhao-Atlas-Marks distributions. Identification methods which employ the aforementioned displays are compared with the more established auditory-acoustic method, using recordings of 10 speakers saying several short phrases two times each. The results point to the efficacy of these newer time-frequency displays for speaker identification, particularly when used to corroborate conventional methods. One important development is the ability to focus on the detailed time-frequency features of the phonation process, thereby obtaining a new speech biometric indicative of vocal cord dynamics.

Audio–video biometric recognition for non-collaborative access granting

In this paper, the problem of non-collaborative person identification for a secure access to facilities is addressed. The proposed solution adopts a face and a speaker recognition techniques. The integration of these two methods allows to improve the performance with respect to the two classifiers. In non-collaborative scenarios, the problem of face recognition first requires to detect the face pattern then to recognize it even when in non-frontal poses. In the current work, a histogram normalization, a boosting technique and a linear discrimination analysis have been exploited to solve typical problems like illumination variability, occlusions, pose variation, etc. In addition, a new temporal classification is proposed to improve the robustness of the frame-by-frame classification. This allows to project known classification techniques for still image recognition into a multi-frame context where the image capture allows dynamics in the environment. For the audio, a method for the automatic speaker identification in noisy environments is presented. In particular, we propose an optimization of a speech de-noising algorithm to optimize the performance of the extended Kalman filter (EKF). To provide a baseline system for the integration with our proposed speech de-noising algorithm, we use a conventional speaker recognition system, based on Gaussian mixture models and mel frequency cepstral coefficients (MFCCs) as features. To confirm the effectiveness of our methods, we performed video and speaker recognition tasks first separately then integrating the results. In particular, two different corpora have been used: (a) a public corpus (ELDSR for audio and FERRET for images) and (b) a dedicated audio/video corpus, in which the speakers read a list of sentences wearing a scarf or a full-face motorcycle helmet. Experimental results show that our methods are able to reduce significantly the classification error rate.

An Evaluation on Recognition Accuracy Rate of Speaker Identification System Using Spectral Ratio between Air-Transmitted Speech Sounds and Bone-Transmitted Speech Sounds

Up to now, a speaker identification method using MFCC spectral ratio between air-transmitted speech sounds and bone-transmitted speech sounds, has been proposed. However, the recognition accuracy rate of this method which is applied to many speakers, has not been shown. This paper shows that the proposed method is useful to identify many speakers using MFCC spectral ratio by the experiment with 70 speakers.

Maximum Likelihood A Priori Knowledge Interpolation-Based Handset Mismatch Compensation for Robust Speaker Identification

Unseen handset mismatch is the major source of performance degradation in speaker identification in telecommunication environments. To alleviate the problem, a maximum likelihood a priori knowledge interpolation (ML-AKI)-based handset mismatch compensation approach is proposed. It first collects a set of handset characteristics of seen handsets to use as the a priori knowledge for representing the space of handsets. During evaluation the characteristics of an unknown test handset are optimally estimated by interpolation from the set of the a priori knowledge. Experimental results on the HTIMIT database show that the ML-AKI method can improve the average speaker identification rate from 60.0% to 74.6% as compared with conventional maximum a posteriori-adapted Gaussian mixture models. The proposed ML-AKI method is a promising method for robust speaker identification.

Combining Gaussianized/Non-Gaussianized Features to Improve Speaker Diarization of Telephone Conversations

We report results on speaker diarization of telephone conversations. This speaker diarization process is similar to the multistage segmentation and clustering system used in broadcast news. It consists of an initial acoustic change point detection algorithm, iterative Viterbi re-segmentation, gender labeling, agglomerative clustering using a Bayesian information criterion (BIC), followed by agglomerative clustering using state-of-the-art speaker identification (SID) methods and Viterbi re-segmentation using Gaussian mixture models (GMMs). We repeat these multistage segmentation and clustering steps twice: once with mel-frequency cepstral coefficients (MFCCs) as feature parameters for the GMMs used in gender labeling, SID, and Viterbi re-segmentation steps and another time with Gaussianized MFCCs as feature parameters for the GMMs used in these three steps. The resulting clusters from the parallel runs are combined in a novel way that leads to a significant reduction in the diarization error rate (DER). On a development set containing 30 telephone conversations, this combination step reduced the DER by 20%. On another test set containing 30 telephone conversations, this step reduced the DER by 13%. The best error rate we have achieved is 6.7% on the development set and 9.0% on the test set.

Integrated speech synthesizer with an automatic identification of speaker connections and identification method used thereof

An integrated speech synthesizer with an automatic identification of speaker connections and identification method includes a sound encode register to store encoded digitized sound data, a first digital to analog converter unit connected to the sound encode register to convert the digitized sound data from the sound encode register to a first analog signal and send out the first analog signal from a first output terminal, a second digital to analog converter unit connected to the sound encode register to convert the digitized sound data from the first output terminal to a second analog signal and send out the second analog signal from the first output terminal and a second output terminal, and a state register connected to the first output terminal to store the state of the first output terminal before the speech synthesizer is enabled. The state stored in the state register is used to set up the initial value and other related settings for the speech synthesizer.

Multistage speaker diarization of broadcast news

This paper describes recent advances in speaker diarization with a multistage segmentation and clustering system, which incorporates a speaker identification step. This system builds upon the baseline audio partitioner used in the LIMSI broadcast news transcription system. The baseline partitioner provides a high cluster purity, but has a tendency to split data from speakers with a large quantity of data into several segment clusters. Several improvements to the baseline system have been made. First, the iterative Gaussian mixture model (GMM) clustering has been replaced by a Bayesian information criterion (BIC) agglomerative clustering. Second, an additional clustering stage has been added, using a GMM-based speaker identification method. Finally, a post-processing stage refines the segment boundaries using the output of a transcription system. On the National Institute of Standards and Technology (NIST) RT-04F and ESTER evaluation data, the multistage system reduces the speaker error by over 70% relative to the baseline system, and gives between 40% and 50% reduction relative to a single-stage BIC clustering system

Robust speaker identification based on selective use of feature vectors

A new method for speaker identification that selectively uses feature vectors for robust decision-making is described. Experimental results, with short speech segments ranging from 0.25 to 2 s, showed that our method consistently outperforms other approaches yielding relative improvements of 20–51% and 15–30% over baseline GMM and the LDA-GMM systems, respectively.

INVESTIGATION OF THE SPEAKER IDENTIFICATION METHOD BASED ON CLUSTERED PSEUDOSTATIONARY SEGMENTS OF VOICED SOUNDS

The problem of speaker identification is investigated. Basic segments ‐ pseudo stationary intervals of voiced sounds are used for identification. The identification is carried out, comparing average distances between an investigative and comparatives. The coefficients of the linear prediction model (LPC) of a vocal tract are used as features of identification. Such a problem arises in stenographic practice where it is important for speech identification to know who is speaking. Identification should be used in stenography and it has to be fast enough in order not to disturb the stenographer's job. The clustered parameter data will be investigated by providing the performance of the speaker identification method with respect to the computational time and the number of errors.

Comparative Study of Speaker Identification Methods: dPLRM, SVM and GMM

A comparison of performances is made of three text-independent speaker identification methods based on dual Penalized Logistic Regression Machine (dPLRM), Support Vector Machine (SVM) and Gaussian Mixture Model (GMM) with experiments by 10 male speakers. The methods are compared for the speech data which were collected over the period of 13 months in 6 utterance-sessions of which the earlier 3 sessions were for obtaining training data of 12 seconds' utterances. Comparisons are made with the Mel-frequency cepstrum (MFC) data versus the log-power spectrum data and also with training data in a single session versus in plural ones. It is shown that dPLRM with the log-power spectrum data is competitive with SVM and GMM methods with MFC data, when trained for the combined data collected in the earlier three sessions. dPLRM outperforms GMM method especially as the amount of training data becomes smaller. Some of these findings have been already reported in [1]--[3].

NOVEL DISCRIMINATIVE VECTOR QUANTIZATION APPROACH FOR SPEAKER IDENTIFICATION

A novel Discriminative Vector Quantization method for Speaker Identification (DVQSI) is proposed, and its parameters selection is discussed. In the training mode of this approach, the vector space of speech features is divided into a number of regions. Then, a Vector Quantization (VQ) codebook for each speaker in each region is constructed. For every possible combination of speaker pairs, a discriminative weight is assigned for each region, based on the region's ability to discriminate between the speaker pair. Consequently, the region, which contains a larger distribution difference between the speech feature vector sets of the two speakers in the speaker pair, plays a more important role by assigning it a larger discriminative weight, in identifying the better speaker match from the two speakers. In the testing mode, to identify an unknown speaker, discriminative weighted average VQ distortion pairs are computed for the unknown speaker input waveform. Then, a technique is described that figures out the best match between the unknown waveform and speakers' templates. The proposed DVQSI approach can be considered a generalization of the existing VQ technique for Speaker Identification (VQSI). The method presented here yields better Speaker Identification (SI) accuracy by employing the discriminative weights and space segmentation as design parameters. This is confirmed experimentally. In addition, a computationally efficient implementation of the DVQSI technique is given which uses a tree-structured-like approach to obtain the codebooks.

홍채와 음성을 이용한 고도의 개인확인시스템

본 논문에서는 홍채인식과 화자식별 방법을 결합한 고도의 개인확인 시스템을 제안한다. 제안된 방법은 홍채인식과 화자식별의 장점만을 부각시켜 최종적으로 결정법칙에 의해 화자를 인식하는 구조로 되어 있으며, 특히 음성에 노이즈가 첨가된 환경하에서도 우수한 성능을 보일 수 있도록 시스템을 구성하였다. 제안된 알고리즘의 성능을 검증하기 위하여 실험한 결과, 보안의 정도가 높은 상급(High)에 해당하는 인식률은 홍채만을 이용한 경우보다 56.7%, 음성을 이용한 화자식별 방법만을 사용한 경우보다 10% 정도 인식률이 향상되었다. 또한 음성에 노이즈를 첨가한 경우에도 보안의 정도가 높은 상급(High)에 해당하는 인식률은 홍채만 이용한 경우보다 인식률이 30%, 노이즈가 첨가된 음성만 이용한 경우보다 60% 정도 향상된 인식률을 보였다. This proposes a new algorithm for advanced personal identification system using iris pattern and speech signal. Since the proposed algorithm adopts a fusion scheme to take advantage of iris recognition and speaker identification, it shows robustness for noisy environments. For evaluating the performance of the proposed scheme, we compare it with the iris pattern recognition and speaker identification respectively. In the experiments, the proposed method showed more 56.7% improvements than the iris recognition method and more 10% improvements than the speaker identification method for high quality security level. Also, in noisy environments, the proposed method showed more 30% improvements than the iris recognition method and more 60% improvements than the speaker identification method for high quality security level.

Method and apparatus for speaker identification using mixture discriminant analysis to develop speaker models

A speaker identification system is provided that constructs speaker models using a discriminant analysis technique where the data in each class is modeled by Gaussian mixtures. The speaker identification method and apparatus determines the identity of a speaker, as one of a small group, based on a sentence-length password utterance. A speaker's utterance is received and a sequence of a first set of feature vectors are computed based on the received utterance. The first set of feature vectors are then transformed into a second set of feature vectors using transformations specific to a particular segmentation unit, and likelihood scores of the second set of feature vectors are computed using speaker models trained using mixture discriminant analysis. The likelihood scores are then combined to determine an utterance score and the speaker's identity is validated based on the utterance score. The speaker identification method and apparatus also includes training and enrollment phases. In the enrollment phase the speaker's password utterance is received multiple times. A transcription of the password utterance as a sequence of phones is obtained, and the phone string is stored in a database containing phone strings of other speakers in the group. In the training phase, the first set of feature vectors are extracted from each password utterance and the phone boundaries for each phone in the password transcription are obtained using a speaker independent phone recognizer. A mixture model is developed for each phone of a given speaker's password. Then, using the feature vectors from the password utterances of all of the speakers in the group, transformation parameters and transformed models are generated for each phone and speaker, using mixture discriminant analysis.

The use of cognitive processing strategies and linguistic cues for efficient automatic language identification

Automatic language identification is clearly an area of research which requires much interdisciplinary effort. However, the major part of the research is carried out by the speech recognition/signal processing community where automatic language identification is seen as being related to speaker independent speech recognition and speaker identification. In particular, speaker identification methods appear to outperform all other methods and the incorporation of prosodic information has contributed only marginally to their success. This is a counterintuitive result that ignores the wealth of data made available by both the linguistic and psycholinguistic communities. The current work presents a methodology based on the cognitive processing strategies which human beings use. In particular, it is proposed that correctly parameterized, prosodic information, such as might be associated with a language with which the listener has little familiarity, is an important discriminating feature. In addition, the hierarchical approach to discrimination which people use, together with suitably parameterized linguistic cues would be more computationally efficient than the indiscriminate extraction and analysis of statistical parameters. In order to illustrate the approach a data set was chosen such as to incorporate examples of what appear to native English speakers to be two quite similar languages and one that is very different from both: Spanish, Portuguese and Chinese. The Hurst exponent is used as a novel parameterization of the prosodic signature resulting from intonation effects over a phrase or short sentence. The results of the illustrative study are in keeping with our own abilities to discriminate between those languages. They indicate that the perception of global, prosodic information subsumes the effect of similarities which may exist between the languages at the local level of acoustic features.

Method and apparatus for speech analysis and synthesis using lattice ladder notch filters

A method and apparatus for analyzing and synthesizing speech includes a programmable lattice-ladder notch filter which may be programmed to exhibit both filter poles and filter zeros and thereby exhibit a power spectral density with a better fit to that of a speech frame such that, when energized by a selected signal sample, a more accurate regeneration of speech is achieved. The filter parameters may be reliably and systematically determined as a single solution to a mathematical analysis given a set of gain parameters matching the observed covariance data and having a prescribed set of transmission zeros. These transmission zeros may either be preselected as a design specification, or recovered from analysis of the speech data. A speech frame may be analyzed and provide a set of parameters which may be transmitted to a remote location where a synthesizer may accurately reproduce the speech frame. A method and apparatus for speaker identification, and speaker verification with a smart card are disclosed implementing the lattice-ladder notch (LLN) filter methodology.

Speaker identification system and method

A speaker identification system (10) employs a supervised training process (100) that uses row action projection (RAP) to generate speaker model data for a set of speakers. The training process employing RAP uses less memory and processing resources by operating on a single row of a matrix at a time. Memory requirements are linearly proportional to number of speakers for storing each speakers information. A speaker is identified from the set of speakers by sampling the speaker's speech (202), deriving cepstral coefficients (208), and performing a polynomial expansion (212) on cepstral coefficients. The identified speaker (228) is selected using the product of the speaker model data (213) and the polynomial expanded coefficients from the speech sample.

Text-independent speaker identification using neural nets and AR-vector models

A method for speaker identification based on the analysis of four sets of speech parameters is proposed. The recognition is carried out by means of two classifiers: the first is based on self-organising Kohonen maps, a new prototype distribution map and a new similarity measure, and the second on AR-vector models. The final decision is realised by a voting principle using the classifiers' decisions.

Speaker identification over telephone system based on channel‐effect cancellation

Abstract This paper presents a method of speaker identification over telephone system based on channel‐effect cancellation. A set of inverse channel spectra is chosen as the basis for generating a channel‐effect‐cancellation filter. The speaker identification scheme is based on vector quantization where the codebooks of each speaker are trained without channel effect. When the speech signal of a speaker goes through the telephone system, a proper channel‐effect‐cancellation filter is generated to minimize the channel effect on the speech signal. The experiments show that the channel effect is mostly contributed by telephone handsets rather than by the transmission lines. By using the proposed method, an identification rate of 93.8% can be achieved for 100 speaker tests.