Abdelilah Jilbab

Discriminating Between Patients With Parkinson’s and Neurological Diseases Using Cepstral Analysis

In this paper, we wanted to discriminate between two groups of patients (patients who suffer from Parkinsons disease and patients who suffer from other neurological disorders). We collected a variety of voice samples from 50 subjects using different recording devices in different conditions. Subsequently, we analyzed and extracted features from these samples using three different Cepstral techniques; Mel frequency cepstral coefficients (MFCC), perceptual linear prediction (PLP), and ReAlitive SpecTrAl PLP (RASTA-PLP). For classification we used leave one subject out validation scheme along with five different supervised learning classifiers. The best obtained result was 90% using the first 11 coefficients of the PLP and linear SVM kernels.

A comparison of skin detection techniques for objectionable videos

In this paper, we propose a real-time system that can categorize input videos into adult or non-adult videos. First, we compare and contrast the most significant skin detection techniques, feature extraction techniques and classification methods. Then, we give an analysis of the significant test results. After careful examination it was decided that an optimal system is gave by a model Bayes-Hsv.

Phonocardiogram signals processing approach for PASCAL Classifying Heart Sounds Challenge

This paper describes a new approach of the first and the second challenge presented by Pattern Analysis, Statistical Modeling and Computational Learning (PASCAL) Classifying Heart Sounds Challenge. The segmentation of phonocardiogram signals into the first heart sound S1 and the second heart sound S2 consists in heart sounds preprocessing, heart sounds peaks detection, extra peaks rejection and S1 and S2 peaks identification. Regarding heart sounds classification into few classes, relevant descriptors have been extracted from phonocardiogram signals, some of which have relied on segmentation results, and used as parameters for an appropriate classifier. The results of this methodology are compared with those of other approaches obtained at PASCAL Classifying Heart Sounds Challenge by means of the segmentation total error value and the precision of each category.

Quantification system of Parkinson's disease

Technological advances in signal processing, electronics, embedded systems and neuroscience have allowed the design of devices that help physicians to better assess the evolution of neurological diseases. In this context, we are interested in the development of an intelligent system for the quantification of Parkinson’s disease (PD). In order to achieve this, the system contains two parts: a wireless sensor network and an embedded system. The wireless sensor network is used to measure motor defects of the patient; it is constituted of several nodes which communicate among themselves. These nodes are intelligent sensors; they contains accelerometers, EMG and blood pressure sensors to detect any malfunction of the patient’s motor activities. As regards to the embedded system, it allows analyzing the patient’s voice signal in order to extract a descriptor that characterizes PD. The network detects the patient’s posture and measures his or her tremors. The voice analysis system measures the degradation of the patient’s condition. The embedded system combines the three decisions using the Chair–Varshney rule. The data fusion between the sensor network and the embedded system, will quantify the disease to facilitate the diagnostic for the physician, while providing the ability to effectively assess the evolution of the patient’s health.

Voice assessments for detecting patients with neurological diseases using PCA and NPCA

In this study, we wanted to discriminate between 30 patients who suffer from Parkinson’s disease (PD) and 20 patients with other neurological diseases (ND). All participants were asked to pronounce sustained vowel /a/ hold as long as possible at comfortable level. The analyses were done on these voice samples. Firstly, an initial feature vector extracted from time, frequency and cepstral domains. Then we used principal component analysis (PCA) and nonlinear PCA (NPCA). These techniques reduce the number of parameters and select the most effective ones to be used for classification. Support vector machine and k-nearest neighbor with different kernels was used for classification. We obtained accuracy up to 88% for discrimination between PD patients ND patients using KNN with k equal to three and five.

FPGA Design and Implementation of an Optimized Adaptive Filter for Real Time Extraction of Vibration Signal Related to Bearing Defects

Self adaptive noise canceller has been successfully used for denoising vibration signal of electrical machine or extracting some of its components. Especially, those related to bearing defects which are crucial for modern maintenance strategy. The high order filter needed for this task requires significant high speed resources. Despite all this, its hardware implementation will reduce both the amount of data to be transmitted and the size of diagnosis applications. This paper investigates the possibility of implementing such filter in Field Programmable Gate Arrays. First, it presents a solution based on an optimized processing unit controlled by a finite state machine. Then, this solution has been improved by paralleling two or three units in order to increase the execution speed, and to meet the needs of the design while providing both chip area and low energy consumption. Finally, the scheduling of this algorithm processes is also presented and timing cost is calculated based on the characteristics and limitations of the hardware. Thus, the different delays are considered in order to ensure real-time processing. The proposed algorithms have been implemented in both fixed and floating point representation.

New approach based on texture and geometric features for text detection

Due to the huge amount of data carried by images, it is very important to detect and identify the text region as accurately as possible before performing any character recognition. In this paper we describe a text detection algorithm in complex background. It is based on texture and connected components analysis. First we abstract texture regions which usually contain text. Second, we segment the texture regions into suitable objects; the image is segmented into three classes. Finally, we analyze all connected components present in each binary image according to the three classes with the aim to remove non-text regions. Experiments on a benchmark database show the advantages of the new proposed method compared to another one. Especially, our method is insensitive to complex background, font size and color; and offers high precision (83%) and recall(73%) as well.