Automatic speech recognition system for Tunisian dialect

Although Modern Standard Arabic is taught in schools and used in written communication and TV/radio broadcasts, all informal communication is typically carried out in dialectal Arabic. In this work, we focus on the design of speech tools and resources required for the development of an Automatic Speech Recognition system for the Tunisian dialect. The development of such a system faces the challenges of the lack of annotated resources and tools, apart from the lack of standardization at all linguistic levels (phonological, morphological, syntactic and lexical) together with the mispronunciation dictionary needed for ASR development. In this paper, we present a historical overview of the Tunisian dialect and its linguistic characteristics. We also describe and evaluate our rule-based phonetic tool. Next, we go deeper into the details of Tunisian dialect corpus creation. This corpus is finally approved and used to build the first ASR system for Tunisian dialect with a Word Error Rate of 22.6%.     

Decision Bayes Criteria for Optimal Classifier Based on Probabilistic Measures

This paper addresses the high dimension sample problem in discriminate analysis under nonparametric and supervised assumptions.Since there is a kind of equivalence between the probabilistic dependence measure and the Bayes classification error probability,we propose to use an iterative algorithm to optimize the dimension reduction for classification with a probabilistic approach to achieve the Bayes classifier.The estimated probabilities of different errors encountered along the different phases of the system are realized by the Kernel estimate which is adjusted in a means of the smoothing parameter.Experiment results suggest that the proposed approach performs well.     

Comparative analysis of RF LDMOS capacitance reliability under accelerated ageing tests

This paper presents the results of comparative reliability study of C–V characteristics through three accelerated ageing tests for stress applied to an RF LDMOS: Thermal shock tests (TST, air–air test), thermal cycling tests (TCT, air–air test) and high temperature storage life (HTSL). The two first tests are carried out with a drain current flowing through the device during stress. The investigation findings of electrical parameter degradations after various ageing tests are discussed. Feedback capacitance (Crs) is reduced by 16% and gate–drain capacitance (Cgd) by 42%. This means that the tracking of these parameters enables to consider the hot carrier injection as the dominant degradation phenomenon. A physical simulation software has been used to confirm qualitatively degradation phenomena.     

A Fittingness Factor-Based Spectrum Management Framework for Cognitive Radio Networks

In order to increase cognitive radios (CRs) operation efficiency, there has been an increasing interest in strengthening awareness level about spectrum utilisation. In this respect, this paper proposes to exploit the fittingness factor concept to capture the suitability of spectral resources exhibiting time-varying characteristics to support a set of heterogeneous CR applications. First, a new knowledge management functional architecture for optimizing spectrum management has been constructed. It integrates a set of advanced statistics capturing the influence of the dynamic radio environment on the fittingness factor. Then, a knowledge manager (KM) exploiting these statistics to monitor time-varying suitability of spectrum resources has been proposed to support the spectrum selection (SS) decision-making process. In particular, a new Fittingness Factor-based strategy combining two SS and spectrum mobility (SM) functionalities has been proposed, following either a greedy or a proactive approach. Results have shown that, with a proper fittingness factor function, the greedy approach efficiently exploits the KM support at low loads and the SM functionality at high loads to introduce significant gains in terms of the user dissatisfaction probability. The proactive approach has been shown to maintain the introduced performance gain while minimizing the signalling requirements in terms of spectrum handover rate.     

An ontology for representing financial headline news

This paper presents the development of an ontology to represent financial headline news. This ontology is developed using the On-To-Knowledge methodology where the focus is on the design steps of the Knowledge Meta Process. This development is part of an ongoing project which aims to design a virtual stock market simulator based on multi-agent systems. The proposed ontology has 31 concepts and includes 201 attributes. The testing results conducted on reliable headline news show that 99% of these headline news can be properly represented by the attributes of the right category in the ontology. Unreliable headline news characterized by news having uncertainties, incompleteness, ill-definition, or imprecision cannot be represented by the proposed ontology. Approaches for representing these unreliable headline news are discussed.     

New pulse mode neuro-fuzzy hardware architecture and application to image denoising

This paper focuses on a new digital architecture of pulse mode neuro-fuzzy system (PMNFS) with on-chip learning ability. The main purpose goal is to make use of the outstanding features of neuro-fuzzy in function approximation, and implement a reconfigurable architecture with on-chip learning on a field-programmable gate array (FPGA) platform. Details of the whole design with on-chip learning solutions are given. As an application illustrating the efficiency and scalability of the proposed PMNFS, we have considered the approximation of image denoising, which is a very important step in image processing. Experimental results show great efficiency of the proposed method, outperforming other denoising techniques. It was also demonstrated that such a system is strongly adaptive and gives good restored images independently of the kind of noises. Owing to learning, such feature cannot be met with conventional denoising techniques. Design synthesis results on a virtex II PRO FPGA platform are presented. Comparisons with conventional techniques as well as neural ones show higher performances of the designed PMNFS.     

A bipolar implementation of the Beta neuron

The great flexibility of the Beta function and its universal approximation characteristics, make Beta basis function neural networks (BBFNNs) very useful. We present a hardware implementation of the Beta neuron. The proposed circuit was designed by using a standard bipolar technology. PSPICE simulations show the good concordance of the output of our circuit with the analytic Beta function. We also successfully integrated the electronic Beta neuron in the design of a BBFNN that approximates a nonlinear mapping.     

Multireference object pose retrieval with volume feedback

Pose retrieval of a rigid object from monocular video sequences or images is addressed. Initially, the object pose is estimated in each image assuming flat depth maps. Shape-from-silhouette is then applied to make a 3-D model (volume), which is used for a new round of pose estimations, this time by a model-based method that gives better estimates. Before repeating this process by building a new volume, pose estimates are adjusted to reduce error by maximizing a novel quality factor for shape-from-silhouette volume reconstruction. The feedback loop is terminated when pose estimates do not change much, as compared with those produced by the previous iteration. Based on a theoretical study of the proposed system, a test of convergence to a given set of poses is devised. Reliable performance of the system is also proved by several experiments on both synthetic and real image sequences. No model is assumed for the object and no feature point is detected or tracked as there is no problematic feature matching or correspondence. Our method can be used for 3-D object tracking in video, 3-D modeling, and volume reconstruction from video.     

Toward best maintenance practices in communications network management

Best maintenance practices in communications networks management are benchmarking standards that, if carefully implemented, will enhance the integrity, reliability and maintenance costs of communications networks. This paper defines best maintenance practices in communications network management within a concise framework encompassing measurable performance‐level goals as well as methods and procedures needed to achieve these goals. The best maintenance practice recommendations of this paper cover many segments of communications networks and services, including wireline and wireless networks as well as multiple network spaces and technologies. The paper also outlines certain recent trends in network maintenance and provides specific recommendations for action to be taken to ensure that best network maintenance practices are implemented and maintained. Copyright © 2005 John Wiley & Sons, Ltd.     

A 90-nm CMOS, 8-bit pipeline ADC with 60-MHz bandwidth and 125-MS/s or 250-MS/s sampling frequency

In this paper a dual operating mode 8-bit, 1.1-V pipeline ADC for Gigabit Ethernet applications is presented. In the two operating modes, the ADC features different sampling frequency (125 and 250 MHz) and power consumption (9.4 and 22.8 mW). Considering a signal bandwidth of 60 MHz in both operating modes, as required by the Gigabit Ethernet standard, the ADC achieves a SNDR always larger than 39.4 dB at 125 MHz and 38.7 dB at 250 MHz (6.25-bit and 6.13-bit ENOB, respectively), with a FoM of 0.84 pJ/conv at 125 MHz and 2.2 pJ/conv at 250 MHz. The ENOB achieved is mainly limited by clock jitter. The ADC is fabricated with a 90-nm CMOS technology, with an active area of 1.25 × 0.65 mm².