Noureddine Ellouze

ECG Beat Classifier Using Support Vector Machine

This paper introduces a new method of heartbeat classification based on the support vector machine classifier using morphological descriptors and High Order Statistic using MIT/BIH Arrhythmia database. Using the morphological descriptors and polynomial kernel, we have obtained an average sensitivity equal to 89,92% and an average specificity about 82,45%, and in the case of Gaussian kernel, we have obtained an average sensitivity equal to 94,26% and an average specificity about 79,02%. Using the High Order Statistic and polynomial kernel, we have obtained an average sensitivity equal to 95,86% and an average specificity about 90,20%, and in the case of Gaussian kernel, we have obtained an average sensitivity equal to 97,15% and an average specificity about 93,07%. The association of the two parameters increases the averages of classification rates; so the sensitivity is 98,38% and the specificity to 94,87% with polynomial kernel and respectively about 94,43% et 95,81 % with Gaussian kernel.

Using robust features with multi-class SVMs to classify noisy sounds

In a sounds recognition system, the most encountered problem is the background noise that can be captured with the sounds to be identified. This paper describes work that has been performed to address this problem. First, the robustness to the environmental noise is investigated for specific kinds of acoustic representation. The representations considered are RASTA-PLP, J-RASTA and wavelets-based processing. Then, we propose to apply multi-class support vector machines (SVMs) as a discriminative framework in order to address audio classification. The experiments conducted on a multi- class problem show that this classifier clearly overperforms the conventional HMM-based system, and hence, we can efficiently address a sounds classification problem characterized by complex real-world datasets, even under important noise degradation conditions.

An Approach Combining Wavelet Transform and Hidden Markov Models for ECG Segmentation

This paper highlights a new method for ECG segmentation based on the combination of two mathematical techniques namely the wavelet transform (WT) and hidden Markov models (HMM). In this method, we first localize edges in the ECG by wavelet coefficients, then, features extracted from the edges serve as input for the HMM. This new approach was tested and evaluated on the manually annotated database QT database (QTDB), which is regarded as a very important benchmark for ECG analysis. We obtained a sensitivity Se= 99,40% for QRS detection and a sensitivity Se= 94,65% for T wave detection.

Speech enhancement based on wavelet packet of an improved principal component analysis

Integrating the principal component analysis in wavelet packet decomposition.Extended PCA technique for speech enhancement is considered.To obtain a sparse matrix to contain the enhanced speech.Experiments on NOIZEUS data corrupted by Gaussian and four non-stationary noises.Our approach shows superior outcomes in BSS EVAL toolbox, SegSNR, PESQ, and Cov. In this paper, we propose a single-channel speech enhancement method, based on the combination of the wavelet packet transform and an improved version of the principal component analysis (PCA). Our method integrates ability of PCA to de-correlate the coefficients by extracting a linear relationship with what of wavelet packet analysis to derive feature vectors used for speech enhancement. This allows us to operate with a convenient shrinkage function on these new coefficients, removing the noise without degrading the speech. Then, the enhanced speech obtained by the inverse wavelet packet transform is decomposed into three subspaces: low rank, sparse, and the remainder noise components. Finally, we calculate the components as a segregation problem. The performance evaluation shows that our method provides a higher noise reduction and a lower signal distortion even in highly noisy conditions without introducing artifacts.

Improved Frame Level Features and SVM Supervectors Approach for The Recogniton of Emotional States from Speech: Application to Categorical and Dimensional States

The purpose of speech emotion recognition system is to classify speakers utterances into different emotional states such as disgust, boredom, sadness, neutral and happiness. Speech features that are commonly used in speech emotion recognition (SER) rely on global utterance level prosodic features. In our work, we evaluate the impact of frame -level feature extraction. The speech samples are fro m Berlin emot ional database and the features extracted fro m these utterances are energy, different variant of mel frequency cepstrum coefficients (MFCC), velocity and accelerat ion features. The idea is to explo re the successful approach in the literature of speaker recognition GMM-UBM to handle with emotion identification tasks. In addition, we propose a classification scheme for the labeling of emotions on a continuous dimensional-based approach. Index Terms—speech emotion recognition, valence, arousal, MFCC, GMM Supervector, SVM

Optimal segments selection for EEG classification

Electroencephalography is non invasive technique used to measure electrical brain activity and to collect EEG signals, through surface electrodes properly positioned on the scalp, which analyzed and interpreted in BCI (Brain Computer Interface) allow the comprehension of human intensions of movement. This is particularly useful for persons with severe physical handicaps. In this paper dataset of motor imagery is used to describe left and right hand movement imagery. To extract discriminating features from EEG signals, wavelet transform is a concrete method which deals with time frequency aspect of non stationary EEG signals. Indeed, the creation of statistical parameters describing wavelet coefficients to build features vectors gives good classification results when used with Linear Discriminate Analysis (LDA) or Neural Networks.

Spectral Analysis of Global Behaviour of C. Elegans Chromosomes

Afef Elloumi Oueslati1, Imen Messaoudi1, Zied Lachiri2 and Noureddine Ellouze1 Unite Signal, Image et Reconnaissance de Formes, Departement de Genie Electrique, 1Ecole Nationale d’Ingenieurs de Tunis, BP 37, Campus Universitaire, Le Belvedere, 1002, Tunis, 2Departement de Genie Physique et Instrumentation Institut National des Sciences Appliquees et de Technologie, BP 676, Centre Urbain Cedex, 1080, Tunis, Tunisie

Detecting particular features in C. elegans genomes using Synchronous Analysis based on Wavelet Transform

In this paper, synchronous analysis based on wavelet transform is applied to genomic sequences. To focus on the particular feature of periodicity 3 in the protein-coding region of genes, a coding method is applied on the sequence, which will be segmented to form a pitch synchronous representation. This modelling concept captures period of period fluctuation of signals. The wavelet transform enhances the periodicity, and a given threshold is used to make decision about exons positions. The algorithm simulation shows the accuracy of the method in simultaneously locating exons and revealing the reading frame associated.

Evaluation of Multi-scale Product Method and DYPSA Algorithm for Glottal Closure Instant Detection

The objective of this paper is to compare two competitive methods of glottal closure instants (GCIs) detection from the speech signal. The two methods are the Multiscale Product (MPM) and Dynamic Programming Phase Slope Algorithm (DYPSA). MPM is based on peaks detection in the product of wavelet transform at three adjacent scales, DYPSA is based on dynamic programming applied on the group delay function. Performance comparison of MPM and DYPSA method is made on Keele University database. We show that GCI estimation by MPM is more precise than DYPSA.

A New Biologically Inspired Fuzzy Expert System-Based Voiced/Unvoiced Decision Algorithm for Speech Enhancement

In this paper, we propose a speech enhancement approach for a single-microphone system. The main idea is to apply a specific transformation on the speech signal depending on the voicing state of the signal. We apply a voiced/unvoiced algorithm based on the multi-scale product analysis with the use of fuzzy logic to make more cognitively inspired use of speech information. A comb filtering is applied on the voiced frames of the noisy speech signal, and a spectral subtraction is operated on the unvoiced frames of the same signal. Further, the harmonics are enhanced by performing a designed comb filtering using an adjustable bandwidth. The comb filter is tuned by an accurate fundamental frequency estimation method. The fundamental frequency estimation method is based on computing the multi-scale product analysis of the noisy speech. Experimental results show that the proposed approach is capable of reducing noise in adverse noise environments with little speech degradation and outperforms several competitive methods.