In this paper, three utterance modelling approaches, namely Gaussian Mean Supervector (GMS), i-vector and Posterior Probability (GPPS), are applied to the accent recognition problem.For each modeling method, different classifiers, Support Vector Machine (SVM), Naive Bayesian Classifier (NBC) Sparse Representation (SRC), employed find out suitable matches between schemes classifiers.The evaluation database is formed by using English utterances of speakers whose native languages Russian,...

Motivated by the success of i-vectors in field speaker recognition, this paper proposes a new approach for age estimation from telephone speech patterns based on i-vectors.In method, each utterance is modeled its corresponding ivector.Then, Support Vector Regression (SVR) applied to estimate speakers.The proposed method trained and tested conversations National Institute Standard Technology (NIST) 2010 2008 Speaker Recognition Evaluations databases.Evaluation results show that outperforms...

We consider a multi-task wireless sensor network (WSN) where some of the nodes aim at applying multi-channel Wiener filter to denoise their local signals, whereas others implementing linearly constrained minimum variance beamformer extract node-specific desired signals and cancel interfering again estimating direction-of-arrival set sources. For this WSN, by relying on distributed signal estimation techniques that incorporate low-rank approximation correlation matrix, we design algorithm...

In many criminal cases, evidence might be in the form of telephone conversations or tape recordings. Therefore, law enforcement agencies have been concerned about accurate methods to profile different characteristics a speaker from recorded voice patterns, which facilitate identification criminal. This paper proposes new approach for gender detection and age estimation, based on hybrid architecture Weighted Supervised Non-Negative Matrix Factorization (WSNMF) General Regression Neural...

Recent studies show that Gaussian mixture model (GMM) weights carry less, yet complimentary, information to GMM means for language and dialect recognition. However, state-of-the-art recognition systems usually do not use this information. In research, a non-negative factor analysis (NFA) approach is developed weight decomposition adaptation. This modeling, which conceptually simple computationally inexpensive, suggests new low-dimensional utterance representation method using similar of the...

Microphone arrays allow to exploit the spatial coherence between simultaneously recorded microphone signals, e.g., perform speech enhancement, i.e., extract a signal and reduce background noise. However, in systems where microphones are not sampled synchronous fashion, as it is often case wireless acoustic sensor networks, sampling rate offset (SRO) exists signals different nodes, which severely affects enhancement performance. To avoid this performance reduction, SRO should be estimated...

We study a distributed node-specific parameter estimation problem where each node in wireless sensor network is interested the simultaneous of different vectors parameters that can be local interest, common interest to subset nodes, or global whole network. assume setting nodes do not know which other share same interests. First, we conduct theoretical analysis on asymptotic bias results case blindly process all estimates their neighbors solve own problem. Next, propose an unsupervised...

This paper proposes a novel approach for automatic estimation of four important traits speakers, namely age, height, weight and smoking habit, from speech signals. In this method, each utterance is modeled using the i-vector framework which based on factor analysis Gaussian Mixture Model (GMM) mean supervectors, Non-negative Factor Analysis (NFA) constrained GMM weights. Then, Artificial Neural Networks (ANNs) Least Squares Support Vector Regression (LSSVR) are employed to estimate height...

This paper proposes a new approach for speaker age estimation.In this method, speakers are modeled by their corresponding Hidden Markov Model (HMM) weight supervectors.Then, Weighted Supervised Non-Negative Matrix Factorization (WSNMF) is applied to reduce the dimension of input space.Finally, Least Squares Support Vector Regressor (LS-SVR) employed estimate using obtained lowdimensional vectors.Evaluation results on corpus read and spontaneous speech in Dutch confirms effectiveness proposed scheme.

This paper proposes a novel approach for automatic speaker height estimation based on the i-vector framework. In this method, each utterance is modeled by its corresponding i-vector. Then artificial neural networks (ANNs) and least-squares support vector regression (LSSVR) are employed to estimate of from given utterance. The proposed method trained tested telephone speech signals National Institute Standards Technology (NIST)2008 2010 Speaker Recognition Evaluation (SRE) corpora...

In this paper, a new fuzzy fading memory (FFM) is developed in order to aid modified input estimation (MIE) technique and enhance its performance tracking high maneuvering targets. The MIE has been introduced recently performs well low medium However, method does not represent desirable accuracy or jerking fact, trouble originated mismodeling the target acceleration dynamics. An effective approach cope with different modeling uncertainties memory. conventional suffers from some deficiencies...

In this paper, a new fuzzy forgetting factor (FFF) is developed in order to aid modified input estimation (MIE) technique and enhance its performance tracking high maneuvering targets. The MIE has been introduced recently succeeds presenting reasonably accurate target trajectory, velocity acceleration low mild situations. However, after some iteration steps become small. Due small steps, the accuracy of may be seriously degraded presence maneuvers. study we present an intelligent self-tuning...

In this paper, a new combined scheme is presented to overcome some drawbacks of the high maneuvering target tracking problems by using mixed fuzzy logic and standard Kalman filter. This consist two important aspects; at first absolute value difference between last course present observation second aspect measurement residual. The results compared with augmented method another which have been reported respectively. Simulation show performance proposed innovation effectiveness in targets problems.

In this paper, a new approach for sampling rate offset (SRO) estimation between nodes of wireless acoustic sensor network (WASN) is proposed using the phase drift coherence function signals. This method, referred to as least squares (LCD) estimation, assumes that SRO induces linearly increasing phase-shift in short-time Fourier transform (STFT) domain. phase-shift, observed signal coherence, applied least-squares framework estimate SRO. Simulation results different scenarios show LCD can...

A distributed multi-speaker voice activity detection (DM-VAD) method for wireless acoustic sensor networks (WASNs) is proposed. DM-VAD required in many signal processing applications, e.g. speech enhancement based on multi-channel Wiener filtering, but non-existent up to date. The proposed neither requires a fusion center nor prior knowledge about the node positions, microphone array orientations or number of observed sources. It consists two steps: (i) source-specific energy unmixing (ii)...

In this paper, we propose a new method for distributed labelling of audio sources in wireless acoustic sensor networks (WASNs). We consider WASNs comprising nodes equipped with multiple microphones observing signals transmitted by sources. An important step toward cooperation between the nodes, e.g. voice-activity-detection, is network-wide consensus on source such that all assign same unique label to each source. hierarchical approach applied first network clustering algorithm performed and...

The task in automatic age recognition speech technology typically is one of regression, i.e., predicting the a speaker from his/her speech. In this paper we are interested probabilistic interpretation posterior distribution predicted age. We review number measures for assessing properties distribution, and link these to detection theory, which very well understood literature. show that Gaussian distributions by least square support vector regression behave well, there only small room...

In this paper a new method based on artificial neural networks (ANN), is introduced for identifying pathogenic antibodies in Systemic Lupus Erythmatosus (SLE). dsDNA binding have been implicated the pathogenesis of autoimmune disease. order to identify these antibodies, protein sequences 42 and 608 non-dsDNA were extracted from Kabat database encoded using physicochemical property their amino acids namely Hydrophilicity. Encoded used as training patterns general regression network (GRNN)....

In this paper, we explored the use of Gaussian Mixture Model (GMM) weights adaptation for speaker verification.We compared two different subspace weight approaches: Subspace Multinomial (SMM) and Non-Negative factor Analysis (NFA).Both techniques achieved similar results seemed to outperform retraining maximum likelihood (ML) adaptation.However, training process NFA approach is substantially faster than SMM technique.The i-vector fusion between each classical yielded slight improvements on...