Speaker Recognition System Research Articles

Voice Disguise in Automatic Speaker Recognition

Humans are able to identify other people’s voices even in voice disguise conditions. However, we are not immune to all voice changes when trying to identify people from voice. Likewise, automatic speaker recognition systems can also be deceived by voice imitation and other types of disguise. Taking into account the voice disguise classification into the combination of two different categories (deliberate/non-deliberate and electronic/non-electronic), this survey provides a literature review on the influence of voice disguise in the automatic speaker recognition task and the robustness of these systems to such voice changes. Additionally, the survey addresses existing applications dealing with voice disguise and analyzes some issues for future research.

GMM and CNN Hybrid Method for Short Utterance Speaker Recognition

During the last few years, the speaker recognition technique has been widely attractive for its extensive application in many fields, such as speech communications, domestics services, and smart terminals. As a critical method, the Gaussian mixture model (GMM) makes it possible to achieve the recognition capability that is close to the hearing ability of human in a long speech. However, the GMM is failing to recognize a short utterance speaker with a high accuracy. Aiming at solving this problem, in this paper, we propose a novel model to enhance the recognition accuracy of the short utterance speaker recognition system. Different from traditional models based on the GMM, we design a method to train a convolutional neural network to process spectrograms, which can describe speakers better. Thus, the recognition system gains the considerable accuracy as well as the reasonable convergence speed. The experiment results show that our model can help to decrease the equal error rate of the recognition from 4.9% to 2.5%.

Genetic Algorithm for Combined Speaker and Speech Recognition using Deep Neural Networks

Huge growth is observed in the speech and speaker recognition field due to many artificial intelligence algorithms being applied. Speech is used to convey messages via the language being spoken, emotions, gender and speaker identity. Many real applications in healthcare are based upon speech and speaker recognition, e.g. a voice-controlled wheelchair helps control the chair. In this paper, we use a genetic algorithm (GA) for combined speaker and speech recognition, relying on optimized Mel Frequency Cepstral Coefficient (MFCC) speech features, and classification is performed using a Deep Neural Network (DNN). In the first phase, feature extraction using MFCC is executed. Then, feature optimization is performed using GA. In the second phase training is conducted using DNN. Evaluation and validation of the proposed work model is done by setting a real environment, and efficiency is calculated on the basis of such parameters as accuracy, precision rate, recall rate, sensitivity, and specificity. Also, this paper presents an evaluation of such feature extraction methods as linear predictive coding coefficient (LPCC), perceptual linear prediction (PLP), mel frequency cepstral coefficients (MFCC) and relative spectra filtering (RASTA), with all of them used for combined speaker and speech recognition systems. A comparison of different methods based on existing techniques for both clean and noisy environments is made as well.

Vocal effort compensation for MFCC feature extraction in a shouted versus normal speaker recognition task

In shouting, speakers use increased vocal effort to convey spoken messages over distance or above environmental noise. For automatic speaker recognition systems trained using normal speech, shouting causes a severe vocal effort mismatch between the enrollment and test hence reducing the recognition performance. In this study, two compensation methods are proposed to tackle the mismatch in a shouted versus normal speaker recognition task. These techniques are applied in the feature extraction stage of a speaker recognition system to modify the spectral envelopes of shouts to be closer to those in normal speech. The techniques modify the all-pole power spectrum of the MFCC computation chain with shouted-to-normal compensation filtering that is obtained using a GMM-based statistical mapping. In an evaluation using the state-of-the-art i-vector based recognition system, the proposed techniques provided considerable improvements in identification rates compared to the case when shouted speech spectra were not processed.

Enhanced Speaker Recognition System

Enhanced Speaker Recognition System

Audio-replay attacks spoofing detection for speaker recognition systems

Audio-replay attacks spoofing detection for speaker recognition systems

Scoring Heterogeneous Speaker Vectors Using Nonlinear Transformations and Tied PLDA Models

Most current state-of-the-art text-independent speaker recognition systems are based on i-vectors, and on probabilistic linear discriminant analysis (PLDA). PLDA assumes that the i-vectors of a trial are homogeneous, i.e., that they have been extracted by the same system. In other words, the enrollment and test i-vectors belong to the same class. However, it is sometimes important to score trials including “heterogeneous” i-vectors, for instance, enrollment i-vectors extracted by an old system, and test i-vectors extracted by a newer, more accurate, system. In this paper, we introduce a PLDA model that is able to score heterogeneous i-vectors independent of their extraction approach, dimensions, and any other characteristics that make a set of i-vectors of the same speaker belong to different classes. The new model, which will be referred to as nonlinear tied-PLDA (NL-Tied-PLDA), is obtained by a generalization of our recently proposed nonlinear PLDA approach, which jointly estimates the PLDA parameters and the parameters of a nonlinear transformation of the i-vectors. The generalization consists of estimating a class-dependent nonlinear transformation of the i-vectors, with the constraint that the transformed i-vectors of the same speaker share the same speaker factor. The resulting model is flexible and accurate, as assessed by the results of a set of experiments performed on the extended core NIST SRE 2012 evaluation. In particular, NL-Tied-PLDA provides better results on heterogeneous trials with respect to the corresponding homogeneous trials scored by the old system, and, in some configurations, it also reaches the accuracy of the new system. Similar results were obtained on the female-extended core NIST SRE 2010 telephone condition.

Performance of speaker recognition system using shifted mfcc, delta spectral cepstral coefficient (DSCC) and Fuzzy techniques

Speech and speaker recognition systems are biometric inspired systems which are having scope in various online and offline applications. In case of biometric we ponder the variability of speech signal due to the presence of noise which greatly degrades the efficiency of Automatic Speaker Recognition (ASR) in real-world environmental circumstances. Real world speech signal is degraded by different types of noise signals like background noise, interference noise and crosstalk noise. In this paper, we have used Delta Spectrum Cepstrum Coefficient (DSCC) and Shifted MFCC with fuzzy modeling techniques to rectify the deed of ASR even in a noisy surrounding with the help of upgraded speech information which is present at high frequency in the spectral domain. The combination of fuzzy modeling and DSCC creates a firm cumulative algorithm which has reasonably high robustness to noise. Experimental results show that accuracy has enhanced by 10-20% even at 5-8dB SNR in the presence of background noise or turbulent environmental condition or in the presence of white noise.Thus proposed model has improved maturity level in comparison to obsolete methods.

Moving average multi directional local features for speaker recognition

A new speech feature extraction technique called moving average multi directional local features (MA-MDLF) is presented in this paper. This method is based on linear regression (LR) and moving average (MA) in the time–frequency plane. Three-point LR is taken along time axis and frequency axis, and 3 points MA is taken along 45° and 135° in the time–frequency plane. The LR captures the voice onset\offset, formant contour, while the moving average captures the dynamics on time–frequency axes which can be seen as voiceprints. The MA-MDLF performance is compared to commonly used speech features in speaker recognition. The comparison is performed in a speaker recognition system (SRS) for three different conditions, namely clean speech, mobile speech, and cross channel. MA-MDLF has shown better performance than the baseline MFCC, RASTA-PLP and LPCC. In clean and mobile speech, MA-MDLF feature performs the best and also in the cross channel task MA-MDLF performed excellent. We also evaluated the MA-MDLF using three speech databases, namely KSU, LDC Babylon and TIMITdatabases, and found that MA-MDLF outperformed the other commonly used features with speech from all the three databases. The first and second databases are for Arabic speech while third is for English speech.

基于多分形谱及特征优选的说话人识别系统

基于多分形谱及特征优选的说话人识别系统

Study of MFCC and IHC Feature Extraction Methods With Probabilistic Acoustic Models for Speaker Biometric Applications

Voice is an important human trait in natural human-to-human interaction / communication for identifying a person. So voice can be regarded as a biometric measure for recognizing or identifying the person similar to other biometric measures such as face, iris and fingerprints. Speaker recognition is a class of voice recognition where speaker is identified from the speech rather than the message. Automatic speaker recognition (SR) is an approach to identify people based on features extracted from speech utterances. The major task in any speaker recognition is to extract useful features and allow meaningful patterns of speaker models. This paper compares the performance of two feature extraction techniques Mel Frequency Cepstral Coefficient (MFCC) and Inner Hair Cell Coefficient (IHC) with two different modelling methods Gaussian Mixture Model - Universal background model (GMM - UBM) and i- vector approach. In this experiment speech samples of 600 speakers from TIMIT database with 10 utterances of each speaker are taken for identifying the speaker. A text independent speaker recognition system was implemented and this study resulted in an inference of MFCC feature outperforms IHC feature for both GMM and i vector. The performance of voiced speech IHC feature which simulates the physiological behaviour of human ear is better in terms of accuracy than full speech (voiced and unvoiced) in GMM.

Концепція впровадження автоматизованої системи розпізнавання мовця у процес автентифікації для доступу до критичної системи

In the article, ISO/IEC 27001: 2013, an effective and user-friendly concept of reliable authentication for accessing a multi-server information critical system, which includes, in particular, a set of users, a set of servers and a dedicated server —a registration center for a system objects registration without the maintenance of verification tables, is proposed. Information exchange between the objects of the system is organized using secure sessions based on an secret keys with a matching mechanism based on one-way hash functions and cryptography of elliptic curves, which now provides the best correlation between encryption reliability and a size of a secret key among existing cryptosystems. Work with users is personalized using individual access cards, which are protected based on the provisions of the cryptographic theory of elliptical curves. Access cards, in addition to identification information and password, store biometric information about the features of the user's voice in the form of a information features vector. The application for authentication, which is installed on the user's computing device, is initiated by a two-step user recognition procedure (with an identification card) — according to the individual features of his voice and the entered password, which together with convenience ensures the reliability of the authentication process. The basic concept of introduction of the automated speaker recognition system in the authentication process for access to the critical system was proposed in the article, and according to the results of the tests it was concluded that it conforms to the requirements of the ISO/IEC 27001: 2013 standard. However, experience in the practical use of the authentication system for accessing information critical system found a number of vulnerabilities in the basic concept, for the purpose of eliminating them, an improved concept was created that also fully meets the requirements of ISO/IEC 27001: 2013. The article evaluates the computational efficiency of the proposed concepts on the basis of calculating the number of the most computationally complex operations — scalable multiplication Tm and calculating the hash function Th for authentication of objects as part of the information critical system (user, server, registration center and all these objects together). The final computational complexity of the basic concept was 4Tm + 15Th, and of the advanced one — 6Tm + 17Th, which in absolute terms is an average result in comparison with analogues. However, the use of cryptography of elliptical curves makes hashing operations about 6 times faster than hashing with the use of a public key, as in analogues.

A SPEAKER RECOGNITION SYSTEM USING GAUSSIAN MIXTURE MODEL

Automatic speaker recognition system identifies a person from the information contained in the speech signal. These systems are the most user-friendly means of biometric recognition and are being used in applications like teleconferencing, banking, forensics etc. The accuracy of these depends on the methods used to extract features from the speech signal, modeling methods, classifiers used to identify the speaker and the size of data available for modeling and verifying. Here, the Mel-Scale Frequency Cepstral Coefficient is one of the methods to grab features from wave file of spoken sentences. The Gaussian Mixture Model is a technique applied in the MARF (Modular Audio Recognition Framework) framework to increase outcome estimation. We have presented a Gaussian selection medium for MFCC.

Review of various stages in speaker recognition system, performance measures and recognition toolkits

Speaker Recognition is a vital application of speech processing. Speaker Recognition performs a task of authenticating or recognizing a speaker based on the unique features captured which characterize the speaker. Characteristics or features which are unique to an individual such as fundamental frequency, speaking style, pitch, and duration are used as distinguishing components of the human speech signal. Exploring these characteristics for various applications with an attempt to implement a robust speaker recognition system has been the impetus behind the research in this domain. This paper makes an attempt to present the available Feature Extraction and Recognition techniques with their merits and demerits. It also discusses the pre-emphasis stage of the speaker recognition system. The standard databases available for speaker recognition along with the criterion for their selection are also reviewed. The paper presents an overview of various toolkits and performance parameters of Automatic Speaker Recognition System.

Binaural Classification-Based Speech Segregation and Robust Speaker Recognition System

The paper presents an auditory scene analyser that comprises of two joint simultaneous modules, namely binaural speech segregation and speaker recognition. The binaural speech segregation is realized by incorporating interaural time and level differences, interaural phase difference and interaural coherence along with direct-to-reverberant ratio into deep recurrent neural network. The performance of deep recurrent network-based speech segregation is validated in terms of source to interference ratio, source to distortion ratio and source to artifacts ratio and compared with existing architectures including deep neural network. It is observed that performance of conventional deep recurrent neural network can be improved further by involving discriminative objectives along with soft time–frequency masking as a layer in the network structure. The system also proposes a spectro-temporal extractor which is referred as Gabor–Hilbert envelope coefficients (GHEC). The proposed monaural feature is responsible for extracting discriminative acoustic information from segregated speech sources. The performance of GHEC is validated under various noisy and reverberant environments and the results are compared with existing monaural features. The results of binaural speech segregation have shown better signal-to-noise ratio at an average of 0.7 dB even in the presence of higher reverberation time, 0.89 s over other baseline algorithms.

A Structured Approach towards Robust Database Collection for Speaker Recognition

<p>Speaker recognition systems are classified according to their database, feature extraction techniques and classification methods. It is analyzed that there is a much need to work upon all the dimensions of forensic speaker recognition systems from the very beginning phase of database collection to recognition phase. The present work provides a structured approach towards developing a robust speech database collection for efficient speaker recognition system. The database required for both systems is entirely different. The databases for biometric systems are readily available while databases for forensic speaker recognition system are scarce. The paper also presents several databases available for speaker recognition systems.</p><p> </p>

Automatic Speaker Recognition: Current Approaches and Progress in Last Six Decades

<p>Automatic speaker recognition is the process to recognizing speaker automatically by their speech/voice on the basis of specific characteristics of his/her speech signal. These voice specific characteristics are called speech features. Over the past six decades many recent advances in the area of speaker recognition have been achieved, but still many problems remains to be solved or require better solutions. The main problems in speaker recognition are session variability, channel mismatch and recording conditions of voice. To develop an efficient speaker recognition system it needs to examine stable parameters of voice features parameters over time, unaffected from variation in speaking, background noise, channel distortion and robust against variation of physical problems. This paper overviews recent advances and general ideas of speaker recognition technology.</p>

Feature Extraction Methods for Speaker Recognition: A Review

This paper presents main paradigms of research for feature extraction methods to further augment the state of art in speaker recognition (SR) which has been recognized extensively in person identification for security and protection applications. Speaker recognition system (SRS) has become a widely researched topic for the last many decades. The basic concept of feature extraction methods is derived from the biological model of human auditory/vocal tract system. This work provides a classification-oriented review of feature extraction methods for SR over the last 55 years that are proven to be successful and have become the new stone to further research. Broadly, the review work is dichotomized into feature extraction methods with and without noise compensation techniques. Feature extraction methods without noise compensation techniques are divided into following categories: On the basis of high/low level of feature extraction; type of transform; speech production/auditory system; type of feature extraction technique; time variability; speech processing techniques. Further, feature extraction methods with noise compensation techniques are classified into noise-screened features, feature normalization methods, feature compensation methods. This classification-oriented review would endow the clear vision of readers to choose among different techniques and will be helpful in future research in this field.

Combining evidences from magnitude and phase information using VTEO for person recognition using humming

Most of the state-of-the-art speaker recognition system use natural speech signal (i.e., real speech, spontaneous speech or contextual speech) from the subjects. In this paper, recognition of a person is attempted from his or her hum with the help of machines. This kind of application can be useful to design person-dependent Query-by-Humming (QBH) system and hence, plays an important role in music information retrieval (MIR) system. In addition, it can be also useful for other interesting speech technological applications such as human-computer interaction, speech prosody analysis of disordered speech, and speaker forensics. This paper develops new feature extraction technique to exploit perceptually meaningful (due to mel frequency warping to imitate human perception process for hearing) phase spectrum information along with magnitude spectrum information from the hum signal. In particular, the structure of state-of-the-art feature set, namely, Mel Frequency Cepstral Coefficients (MFCCs) is modified to capture the phase spectrum information. In addition, a new energy measure, namely, Variable length Teager Energy Operator (VTEO) is employed to compute subband energies of different time-domain subband signals (i.e., an output of 24 triangular-shaped filters used in the mel filterbank). We refer this proposed feature set as MFCC-VTMP (i.e., mel frequency cepstral coefficients to capture perceptually meaningful magnitude and phase information via VTEO)The polynomial classifier (which is in-principle similar to other discriminatively-trained classifiers such as support vector machine (SVM) with polynomial kernel) is used as the basis for all the experiments. The effectiveness of proposed feature set is evaluated and consistently found to be better than MFCCs feature set for several evaluation factors, such as, comparison with other phase-based features, the order of polynomial classifier, person (speaker) modeling approach (such as, GMM-UBM and i-vector), the dimension of feature vector, robustness under signal degradation conditions, static vs. dynamic features, feature discrimination measures and intersession variability.

DEVELOPMENT OF VOICE ACTIVATED SPEAKER RECOGNITION SYSTEM.

20Jul 2017 DEVELOPMENT OF VOICE ACTIVATED SPEAKER RECOGNITION SYSTEM. Chaitra K N , Anjan Kumar B S and H N Suresh. M.tech Student, Department of E&I, Bangalore Institute of Technology, Bengaluru. Assistant professor, Department of E&I, Bangalore Institute of Technology, Bengaluru. PG Coordinator, Department of E&I, Bangalore Institute of Technology, Bengaluru.