A. Djeradi

• Professor
• University of Sciences and Technology Houari Boumediene

This work conducts a comparative investigation of two architectures in the domain of Spoken Language Understanding (SLU), which were evaluated on a synthesized corpus of three languages: Modern Standard Arabic (MSA), French, and English. The first architecture employs a simple SLU system based on classical machine learning algorithms (E2E SLU), whe...

Static hand gesture (HG) recognition for both user-dependent and user-independent is a challenging problem, especially when there are changes in lighting, hand position, and background, the recognition becomes more complex. To solve this problem, this paper proposes a static hand gesture recognition based on a set of image descriptors: Gradient Loc...

As the most used approach to extend a Spoken language Understanding (SLU) from a language to another, Machine translation achieves high performance for English domains, which is not the case for other languages, especially low-resourced ones as Arabic and its dialects. To avoid Machine Translation approach which requires huge parallel corpora, we w...

In this paper, we suggest the generalization of an Arabic Spoken Language Understanding (SLU) system in a multi-domain human-machine dialog. We are interested particularly in domain portability of SLU system related to both structured (DBMS) and unstructured data (Information Extraction), related to four domains. In this work, we used the thematic...

As the most used approach to extend a Spoken language Understanding (SLU) from a language to another, Machine translation achieves high performance for English domains, which is not the case for other languages, especially low-resourced ones as Arabic and its dialects. To avoid Machine Translation approach which requires huge parallel corpora, we w...

Audiovisual data compression has received a considerable progress in recent years. A lot of research focused on the application of the Discrete Cosine transform DCT and the Discrete wavelet transform DWT for JPEG and MPEG compression, while the LPC model has been widely used for the acoustic data compression. The aim of our work is to study the eff...

In the present paper, we suggested an implementation of an automatic understanding system of the statement in Human-Machine Communication. The architecture we adopted was based on a stochastic approach that assumes that the understanding of a statement is nothing but a simple theme identification process. Therefore, we presented a new theme identif...

In the present paper, we suggest an implementation of an automatic understanding system of the statement in human-machine communication. The architecture we adopt is based on a stochastic approach that assumes that the understanding of a statement is nothing but a simple theme identification process. Therefore, we present a new theme identification...

The objective of this work is to investigate the benefit of discrete wavelet transform combined with LPC, for speaker identification system applied for Algerian Berber language, compared to the traditional Mel frequency analysis. We’ve developed a speaker identification system for Algerian Berber language. The corpus concerns two dataset, the first...

This paper presents a method of tracking points on a speaking face and the reconstruction of a 2D face model from speaking descriptive vectors followed. After capturing a video of a talking face using a CCD camera, his facial expression changes in a video sequences. The essential of this work is to track the feature points on the face using the Luc...

Our study consists of a phenomena analysis involved in the production of constrained speech. The paradigm used is the variation of speech rate. A variable rate flow speech simulation study was performed using the ISHIZAKA vocal tract mechanical model (two-mass model) grafted on the Klatt formant synthesizer (acoustic speech production model). enabl...

In the present paper, we suggested an implementation of an automatic understanding system of the statement in Human-Machine Communication. The architecture we adopted was based on a stochastic approach that assumes that the understanding of a statement is nothing but a simple theme identification process. Therefore, we presented a new theme identif...

Dans le présent papier, nous proposons la mise en oeuvre d’un système de Compréhension Automatique de l’énoncé en Communication Homme-Machine. L’architecture que nous adoptons est basée sur une approche Statistico-Linguistique qui suppose que la compréhension d’un énoncé n’est autre qu’une traduction du langage naturel vers le langage conceptuel. P...

This paper deals with the underlining and the prominent display of the compensatory strategies developed by six speakers which are from different geographic regions when they express themselves in a second language very different from their mother native language. The paradigm that had been used in this study is the elocution speed as a speaking co...

This paper presents a method of tracking points on a speaking face and the reconstruction of a face model from speaking descriptive vectors followed. After capturing a video of a talking face using a CCD camera, his facial expression changes in a video sequences. The essencial of this work is to track the feature points on the face using the Lucas-...

We used locus equations for characterizing the two berber consonants 'lip-vélarized' /gw/and /kw/. The aim is to show that these two phonemes are consonants distinct from their homologous velar /g/and /k/. The second and third order of locus equations have produced appreciable results

One of the major issues when transforming a voice using the PSOLA algorithm is to be able to accurately find the values for the signal modification parameters (α, β and γ) that allow us to transform the source signal into the target signal. In this paper, we propose a way to determine these parameters on the basis of a study of their influence on s...

The acoustic measurement of articulation place for consonants and the acoustic measurement of the amount of coarticulation are two long-standing problems in phonetics. Previous work on consonant place of articulation, using articulatory-acoustic models, within the acoustic theory of speech production (Stevens, 1998), has found certain acoustic cor...

Parmi les techniques appliquées au domaine de la Traitement Automatique du Langage Naturel TALN, on y trouve ceux appartenant à l'approche linguistique. Cette dernière tient en considération les informations linguistiques liées au corpus de l'application visée. Ces informations nous aident à concevoir un système de compréhension d'énoncé dans un di...

Our study consists of analysing Arabic utterances VCVα in brief vocalic context with Vα and speech rate as variables in order to observe the impact of the "right" context and speech rate on the coarticulation. Thus, we have to look for some invariance in the speech signal explaining the coarticulation phenomenon related to speech rate. So, we have...

This study aims to identify the compensatory strategies that speaker develops in speech production in noisy environment. We have developed a method which allows studying the effect of noise on acoustic parameters that characterize the speech signal (F0 duration, formants, cepstral and LPC parameters). The results reflect the speakers' attitudes tow...

We used the locus equations and formant trajectories to characterize two consonants " lip-vélarized " / gw / and / kw /. We have proposed a method of representing the formant trajectory by the slope and the intercept of the linear regression. This method allowed to provide information to complement locus equations and helped distinguish between " l...

Visemes are the unique facial positions required to produce phonemes, which are the smallest phonetic unit distinguished by the speakers of a particular language. Each language has multiple phonemes and visemes, and each viseme can have multiple phonemes. However, current literature on viseme research indicates that the mapping between phonemes and...

Th is article p resents a robust method for detecting iris features in frontal face images based on circular Hough transform. The software of the applicat ion is based on detecting the circles surrounding the exterio r iris pattern from a set of facial images in d ifferent color spaces. The circular Hough transform is used for this purpose. First a...

The labiovelarization of velar consonants and labials is a very widespread phenomenon. It is attested in all the major northern Berber dialects. Only the Tuareg completely ignores it. But, even within the large Berber-speaking regions of the north, it is very unstable: it may be completely absent in some languages (such as the Bougie region in Kaby...

Proposed is an efficient face recognition algorithm using the discrete cosine transform DCT Technique for reducing dimensionality and image parameterization. These DCT coefficients are examined by a MLP Multi-Layer Perceptron and radial basis function RBF neural networks. Their purpose is to present a face recognition system that is a combination o...

We propose in this study, a method allowing us to characterize a speech occurred in a stress situation. For this, we created an artificial disturbance (stress lip) and then, we analyzed the effects of stress on the acoustic parameters of the signals produced. We have developed a methodology allowing us to analyze the timing, the fundamental frequen...

This study is part of adaptive mechanisms research in speech production. We were interested to acoustic variations of the voice message when the speaker is placed in a noisy environment. We propose a methodology to highlight the articulatory strategies adopted by the speaker to adapt to the noisy environment. This methodology consists of two parts:...

This article develops a speaker-dependent Arabic phonemes recognition system using MFCC analysis and the VQ-LBG algorithm. The system is examined with and without vector quantization in order to analyze the effect of compression in an acoustic parameterization phase. Our experimental results show that vector quantization using a codebook of size 16...

We acoustically analyzed behavior of speech signal produced in noisy constraint by four speakers, when noise is sent by a helmet to speaker and when noise is sent by high speaker. The goal is to find speech signal acoustic parameters which are most sensitive to noise and the compensatory strategies adopted by the speakers to counter this constraint...

Visemes are the unique facial positions required to produce phonemes, which are the smallest phonetic unit distinguished by the speakers of a particular language. Each language has multiple phonemes and visemes, and each viseme can have multiple phonemes. However, current literature on viseme research indicates that the mapping between phonemes and...

Speaker recognition is the process of automatically recognizing who is speaking on the basis of individual information included in speech waves. Feature extraction for speech recognition is a subject of a major interest today; different features have been investigated in speech recognition systems. The perceptual linear predictive PLP: this techniq...

The goal of this study is to consider the instantaneous frequencies corresponding to the speech signal formant using the wavelet transform. The developed method is based on an analysis of derivative phase of the continuous Morlet wavelet transform coefficients. Using synthesized signals produced by a formant model made it possible to adapt this met...

We present a fast method for locating iris features in frontal face images based on the Hough transform. The aim of this work is to detect the circles surrounding the exterior iris pattern from a set of facial images. The circular Hough transform is used for this purpose, a first edge detection technique is used for finding the edges in the input i...

Audio-only speaker/speech recognition systems ASR are far from being perfect especially under noisy conditions. Furthermore, it is a known fact that the content of speech can be revealed partially through lip-reading. Human speech perception is bimodal in nature: Humans combine audio and visual information in deciding what has been spoken, especial...

Knowledge of vocal tract area functions is important for the understanding of phenomena occurring during speech production. We present here a new measurement method based on the external excitation of the vocal tract with a known pseudo-random sequence, where the area function is obtained by a linear prediction analysis applied to the cross-correla...

In this paper, we present a face recognition system that is a combination of cosinus discrete transform algorithm with a neural network. Discrete cosine transforms is used to reduce image information redundancy because only a subset of the transform coefficients are necessary to preserve the most important facial features. Multilayer perceptrons (M...

This study treats the effects of speech rate in a second language. Six speakers from different geographical areas have produced sentences in written Arabic, carrying fricatives (Arabic specifications), at different speeds of elocution. The selected speakers are: two Lebanese (CH and LI), two inhabitants of Algiers (FE and MA) and two Kabyles (SAand...

Nous présentons une étude paramétrique d’un modèle de la source vocale, connu comme modèle à deux masses. Ce modèle nous a permis de caractériser la fréquence fondamentale de la source en fonction de la pression des poumons et la tension des cordes vocales. Les résultats d’une telle étude contribuent à une meilleure connaissance de la réponse d’un...

The paper presents a modelling of the human machine dialogue in dual language (French and Arabic). According to the principle of our model the production of a speech result from an exchange between the human and the machine, that depends on the interlocutor's mental condition. It presents itself under the set of fundamental units of the memory (FUM...

Face is the most common biometric identifier used by humans. During the past thirty years, a number of face recognition techniques have been proposed, all of these methods focus on image-based face recognition that use a still image as input data. In this paper, Linear Discriminant Analysis (LDA) which is also called fisherface is an appearance-bas...

The degree of coarticulation and the vocalic reduction (RV) are indices related to a good engine control (Gay, 1978). Fowler (1998) explains why locus equation (LE) is used to characterize, at the same time, the place of articulation and the degree of coarticulation between consonants and vowels: a strong slope (m=1) indicates a maximum coarticulat...

The degree of coarticulation and the vocalic reduction (RV) are indices related to good engine control (Gay 1978). Fowler (1998) explains why locus equation (LE) is used to characterize, at the same time, the place of articulation and the degree of coarticulation between consonants and vowels: a strong slope (m=1) indicates a maximum coarticulation...

The determination of the glottic source parameters is a relatively difficult subject, because it relates to the measurement of parameters convoluted of the vocal source. If, using a recording EGG, it is possible to reach two of the principal parameters of analysis, there is not currently reliable method to determine the remaining parameters. Thus,...

Locus equations are linear regressions of the onset of F2 transitions on their offset .These functions are able to characterise consonantal place categories. Taking up again previous literature studies; this experiment explored the information for place of articulation provided by these functions to the two Arabic plosive /q/ (/‫ﻖ‬ /) and /?/ (/‫)/...

Locus equations are linear regressions of the onset of F2 transitions on their offset. These functions are able to characterize consonantal place categories. This experiment explored the accuracy in the information for place of articulation provided by these functions to the two Arabic plosive /q/ (/‫/ﻖ‬) and /?/ (/‫)/ﺀ‬ and their corresponding fri...

Locus equations are linear regressions of the onset of F2 transitions on their offset .These functions are able to characterise consonantal place categories. Taking up again previous literature studies; this experiment explored the information for place of articulation provided by these functions to the two Arabic plosive /q/ (/‫ﻖ‬ /) and /?/ (/‫)/...

The determination of the glottic source parameters is a relatively difficult subject, because it relates to the measurement of parameters convoluted of the vocal source. If, using a recording EGG, it is possible to reach two of the principal parameters of analysis, there is not currently reliable method to determine the remaining parameters. Thus,...

The determination of the glottic source parameters is a relatively difficult subject, because it relates to the measurement of parameters convoluted of the vocal source. If, using a recording EGG, it is possible to reach two of the principal parameters of analysis, there is not currently reliable method to determine the remaining parameters. Thus,...

Speech coders operating at low bit rates necessitate efficient encoding of the linear predictive coding (LPC) coefficients. Line spectral frequencies (LSF) parameters are currently one of the most efficient choices of transmission parameters for the LPC coefficients. In this paper, an optimized trellis coded vector quantization (TCVQ) scheme for en...

Un des problèmes importants dans le codage de la parole à bas débit est la conception de quantificateurs efficaces pour le codage des coefficients de prédiction linéaire (Lpc). Les paramètresLsf (Line spectral Frequencies) sont actuellement classés parmi les choix les plus appropriés pour représenter les coefficientsLpc. Dans cet article, un systèm...

Wave propagation in a lossy vocal-tract with yielding walls is studied. Simulation method of this vocal-tract is described. The method adopted allowed us to calculate the transfer function, the formant frequencies and the formant bandwidths of the vocal-tract. The results obtained permit the determination of the differential contribution of the dif...

Speech coders operating at low bit rates necessitate efficient encoding of the linear predictive coding (LPC) coefficients. Line spectral Frequencies (LSF) parameters are currently one of the most efficient choices of transmission parameters for the LPC coefficients. In this paper, an optimized trellis coded vector quantization (TCVQ) scheme for en...

Knowledge of vocal tract area functions is important for the understanding of phenomena occurring during speech production. We present here a new measurement method based on the external excitation of the vocal tract with a known pseudo-random sequence, where the area function is obtained by a linear prediction analysis applied to the cross-correla...

Trellis coded quantization, both scalar and vector, improves upon traditional trellis encoded systems by labelling the trellis branches with entire subsets rather than with individual reproduction levels. Trellis Coded Vector Quantization (TCVQ) was introduced as an effective low-complexity source coding technique which achieves rate-distortion per...

Knowledge of vocal tract area function is important for the understanding of phenomena occurring during speech production. We present here a new measurement method based on external excitation of the vocal tract with a known pseudo-random sequence, where the area function is obtained by a linear prediction analysis applied at the cross-correlation...

Distortion of parameters of a vector quantization system can be alleviated by using an index assignment channel-source coding technique. To carry out this approach, we have developed an iteratif algorithm based on the principle of the simulated Annealing. The goal of this algorithm is to find globally optimal channel-codes, destined for the implici...

The paper presents a modelling of the human machine dialogue in dual language (French and Arabic). According to the principle of our model the production of a speech result from an exchange between the human and the machine, that depends on the interlocutor's mental condition. It presents itself under the set of fundamental units of the memory (FUM...