Meriem Jaidane

Non-Negative Blind Source Separation Algorithm Based on Minimum Aperture Simplicial Cone

We address the problem of Blind Source Separation (BSS) when the hidden sources are Nonnegative (N-BSS). In this case, the scatter plot of the mixed data is contained within the simplicial cone generated by the columns of the mixing matrix. The proposed method, termed SCSA-UNS for Simplicial Cone Shrinking Algorithm for Unmixing Non-negative Sources, aims at estimating the mixing matrix and the sources by fitting a Minimum Aperture Simplicial Cone (MASC) to the cloud of mixed data points. SCSA-UNS is evaluated on both independent and correlated synthetic data and compared to other N-BSS methods. Simulations are also performed on real Liquid Chromatography-Mass Spectrum (LC-MS) data for the metabolomic analysis of a chemical sample, and on real dynamic Positron Emission Tomography (PET) images, in order to study the pharmacokinetics of the [18F]-FDG (FluoroDeoxyGlucose) tracer in the brain.

Blind cancellation of intersymbol interference in decision feedback equalizers

High speed data transmission over bandlimited and/or dispersive channels require equalization to remove the intersymbol interference (ISI) between the successive transmitted data. This paper presents the cancellation of intersymbol interferences (ISI) as a criterion for the blind adaptation of decision feedback equalizers (DFE). We show that this criterion is an alternative to the decision directed (DD) algorithm. This paper also proposes to replace the hard limiter in the decision device of the DFE by a soft decision implemented by a hyperbolic tangent function in order to escape from local minima during the initialization of the blind algorithms. As the theoretical investigation of these criteria and algorithms is difficult, we analyse some simple examples to illustrate the interest of the proposed solutions.

TREND EXTRACTION FOR SEASONAL TIME SERIES USING ENSEMBLE EMPIRICAL MODE DECOMPOSITION

In this paper, we investigate eligibility of trend extraction through the empirical mode decomposition (EMD) and performance improvement of applying the ensemble EMD (EEMD) instead of the EMD for trend extraction from seasonal time series. The proposed method is an approach that can be applied on any time series with any time scales fluctuations. In order to evaluate our algorithm, experimental comparisons with three other trend extraction methods: EMD-energy-ratio approach, EEMD-energy-ratio approach, and the Hodrick–Prescott filter are conducted.

Nonlinear Audio Systems Identification Through Audio Input Gaussianization

Nonlinear audio system identification generally relies on Gaussianity, whiteness and stationarity hypothesis on the input signal, although audio signals are non-Gaussian, highly correlated and non-stationary. However, since the physical behavior of nonlinear audio systems is input-dependent, they should be identified using natural audio signals (speech or music) as input, instead of artificial signals (sweeps or noise) as usually done. We propose an identification scheme that conditions audio signals to fit the desired properties for an efficient identification. The identification system consists in (1) a Gaussianization step that makes the signal near-Gaussian under a perceptual constraint; (2) a predictor filterbank that whitens the signal; (3) an orthonormalization step that enhances the statistical properties of the input vector of the last step, under a Gaussianity hypothesis; (4) an adaptive nonlinear model. The proposed scheme enhances the convergence rate of the identification and reduces the steady state identification error, compared to other schemes, for example the classical adaptive nonlinear identification.

Ultrasound Backscatter Characterization by Using Markov Random Field Model

This paper evaluates a K-Markov random field model for retrieving information about backscatter characteristics, especially regularity spacing scatterers in simulated ultrasound image. The model combines a statistical K-distribution that describes the envelope of backscattered echo and spatial interaction given by Markov random field (MRF). Parameters estimated by the conditional least squares (CLS) estimation method on simulated radio-frequency (RF) envelope image show that the interaction parameters measure the degree of the randomness of the scatterers

Doping and witness watermarking for audio processing

In the last years the research in the field of digital watermarking evolved toward non conventional watermarking applications, which we refer to as watermark aided audio processing. In this paper, we present some of our contributions to this new concept of watermarking for audio processing, namely doping watermarkand witness watermark. The doping watermarking modifies the statistical characteristics of signals in order to fulfill particular conditions requested by algorithms or systems: it may stationarize, gaussianize a signal or make its characteristic function band-limited. We further show that a watermark can be used as a witness to channel distortions, and hence facilitates channel identification.

Exact performances analysis of a selective coefficient adaptive algorithm in acoustic echo cancellation

In the hands-free communications, the identification of long impulse response in acoustic echo cancellation requires very important load calculations. One way to reduce the complexity of the classical normalized least mean square (NLMS) adaptive algorithm, is to use the Mmax NLMS algorithm (Aboulnasr and Mayyas 1999). It is shown that this algorithm is a very promising one, that maintains closest performance to the full update NMLS filter in spite of the updating of a small number of coefficients. However, due to its complexity, the mean square analysis uses an unrealistic hypothesis. It was then not possible to consider the practical context such as high input correlation or high step size. In this paper, we present an exact performances analysis, inspired from Besbes et al. (1999), when the input signal remains in a finite alphabet set. With this realistic hypothesis, dedicated to the digital context, we can describe accurately the Mmax NLMS behavior without any unrealistic assumption. In particular, we evaluate the exact value of critical and optimal step size and we provide the exact mean square error (MSE) for all step sizes and input correlations. The influence of high order statistics can be enhanced.

Arabic adaptation of Phonology and Memory test using entropy-based analysis of word complexity

Short-term Phonological Memory evaluation is very important in tracking the learning development of children, and for that, speech-therapists use words with different linguistic complexity levels. The test “Memory and Phonology” (PM) adopted in the Department of Child and Adolescent Neurology in the National Institute of Neurology (Tunisia), relies on a heuristically-constructed database of Tunisian-Arabic words, in analogy with the standardized NEEL PM French test. Given that the choice of a words phonology is very specific to the language and a costly experimental calibration must be done to validate the classification according to words levels difficulty, this work offers a phonology-based complexity algorithm that validates not only the French data base of the NEELs test, but also that currently proposed and adopted by the Tunisian speech therapists. The proposed complexity analysis and classification are based on three entropy measures: time-based, spectral, and the variance of dynamic spectral entropy of the sound signal of a given word.

AffectiveROAD system and database to assess driver's attention

Thanks to the rise of new wearable and non-intrusive sensor technology, Internet of Things (IoT) contributes in human daily life improvement. In the context of smart vehicles, human affective monitoring should be based on a context-aware system in order to consider the interactions between the driver, the vehicle and the ambient environment. In this paper, we propose AffectiveROAD platform, that senses the human physiological changes, the ambient environment inside the vehicle, and the vehicle speed. Thanks to this platform, several drivers state indicators such as stress and arousal may be developed and validated. Two types of wireless physiological sensors are used to monitor the electrodermal activity, the heart rate, the skin temperature, the respiration, and the hand movement of the driver. Moreover, we developed a sensor network allowing to capture the ambient temperature, humidity, pressure, and luminosity. The purpose of this paper is to describe a real-world driving protocol allowing to collect data using IoT-based materials and to announce the publication of a database for drivers state monitoring research. A partial database concerning the physiological and the environmental information is available on request, for public use.

Speech based transmission index for all: An intelligibility metric for variable hearing ability

A method to measure the speech intelligibility in public address systems for normal hearing and hearing impaired persons is presented. The proposed metric is an extension of the speech based Speech Transmission Index to account for accurate perceptual masking and variable hearing ability: The sound excitation pattern generated at the ear is accurately computed using an auditory filter model, and its shapes depend on frequency, sound level, and hearing impairment. This extension yields a better prediction of the intensity of auditory masking which is used to rectify the modulation transfer function and thus to objectively assess the speech intelligibility experienced by hearing impaired as well as by normal hearing persons in public spaces. The proposed metric was developed within the framework of the European Active and Assisted Living research program, and was labeled SB-STI for All. Extensive subjective in-Lab and in vivo tests have been conducted and the proposed metric proved to have a good correlation with subjective intelligibility scores.