By Topic

Intelligence and Systems, 1998. Proceedings., IEEE International Joint Symposia on

Date 23-23 May 1998

Filter Results

Displaying Results 1 - 25 of 61
  • Proceedings. IEEE International Joint Symposia on Intelligence and Systems (Cat. No.98EX174)

    Save to Project icon | Request Permissions | PDF file iconPDF (330 KB)  
    Freely Available from IEEE
  • Author index

    Page(s): 429 - 430
    Save to Project icon | Request Permissions | PDF file iconPDF (155 KB)  
    Freely Available from IEEE
  • Recognition of Arabic phonetic features using neural networks and knowledge-based system: a comparative study

    Page(s): 404 - 411
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (144 KB)  

    This paper deals with a new indicative features recognition system for Arabic which uses a set of a simplified version of sub-neural-networks (SNN). For the analysis of speech, the perceptual linear predictive technique is used. The ability of the system has been tested in experiments using stimuli uttered by 6 native Algerian speakers. The identification results have been confronted to those obtained by the SARPH knowledge based system. Our interest goes to the particularities of Arabic such as geminate and emphatic consonants and the duration. The results show that SNN achieved well in pure identification while in the case of phonologic duration the knowledge-based system performs better View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Visible speech modelling and hybrid hidden Markov models/neural networks based learning for lipreading

    Page(s): 336 - 342
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (48 KB)  

    This paper describes a new approach for automatic visible speech recognition based on hybrid hidden Markov models/neural networks. Suitable geometric features extracted from speaker's lip shapes are used to train the speech recognizer with nonsense sentences. First we describe the use of a geometrical-based model for visible speech and we outline a self-organising-map-based approach to determine the visual specific recognition units suitable for our speaker-dependent visible speech recognition task. Then we describe five automatic lipreading systems we developed according to different classification techniques: hidden Markov models, neural networks and hybrid hidden Markov models/neural networks. All these systems are tested on a connected letter recognition task. Finally, the performance comparison underlines that a hybrid hidden Markov models/neural networks based architecture is the most promising for automatic lipreading purposes View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Shape recognition and orientation detection for industrial applications using ultrasonic sensors

    Page(s): 301 - 308
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (88 KB)  

    This paper deals with a method for recognizing the form and orientation of pieces. This system uses a single pair of ultrasonic sensors to distinguish different objects and their orientations, for a set of previously learned objects. This technique utilizes the feature that small variations of position produce small variations in the value of the echo envelope parameters characterizing the ultrasonic signal. Then, neural nets are applied to learn and retrieve the necessary data in order to obtain the real position of the object. Several NN structures have been tested in order to find those that provide the best results. This system has been evaluated with symmetrical geometrical figures. Subsequently, the application was utilized in a robotic system View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural techniques for image segmentation

    Page(s): 367 - 372
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (104 KB)  

    We present new neural techniques including unsupervised technology and fuzzy logic foundations. We realized a hybrid neural network and applied three different unsupervised learning algorithms that we developed specially for it: fuzzy MLSOM, fuzzy hierarchical “neural gas” and fuzzy hierarchical “maximum entropy”. The experiments presented deal with image segmentation. The results obtained show that neural networks are a valid instrument for image processing and shape recognition View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Algorithms for splicing junction donor recognition in genomic DNA sequences

    Page(s): 169 - 176
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (48 KB)  

    The consensus sequences at splicing junctions in genomic DNA are required for pre-mRNA breaking and rejoining which must be carried out precisely. Programs currently available for identification or prediction of transcribed sequences from within genomic DNA are far from being powerful enough to elucidate genomic structure completely. In this paper, we develop pattern matching algorithms for 5' splicing site (donor site) recognition. Using the Motif models we develop, we can extract the degenerate pattern information from the consensus splicing junction sequences. The experimental results show that, our algorithm could correctly recognize 93% of the total donor sites at the right positions in the test DNA group. Furthermore, more than 91% of the donor sites were correctly predicted by our algorithm. These precision rates are higher than the best existing donor classification algorithm View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Directing the structure of matter through DNA nanotechnology

    Page(s): 146 - 150
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (64 KB)  

    The sticky-ended association of DNA molecules occurs with high specificity, and it results in the formation of B-DNA, whose structure is well known. The use of stable branched DNA molecules permits one to make stick-figures. We have used this strategy to construct a covalently closed DNA molecule whose helix axes have the connectivity of a cube, and a second molecule, whose helix axes have the connectivity of a truncated octahedron. In addition to branching topology, DNA also affords control of linking topology, because double helical half-turns of B-DNA or Z-DNA can be equated, respectively, with negative or positive crossings in topological objects. Consequently, we have been able to use DNA to make trefoil knots of both signs and figure-8 knobs. DNA-based topological control has also led to the construction of Borromean rings. The key feature previously lacking in DNA construction has been a rigid molecule. We have discovered that antiparallel DNA double crossover molecules can provide this capability View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Three tier architecture for controlling space life support systems

    Page(s): 195 - 201
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (32 KB)  

    Managing life support for remote facilities requires maintaining environmental conditions beneficial to humans and plants, and managing resources like water, product gases, and food. Appropriate allocation and coordination of these tasks among humans, robots, and life support systems is important for efficient operations. The need for operational flexibility and reactivity combined with the need to reduce crew workload and the high cost of failure to manage resources effectively suggests automating low level control tasks and assisting humans in strategic planning and resource management. The three tier (3T) layered architecture is well-suited for such automated control. The planner automates task coordination across subsystems contending for resources. The separation of deliberative tasks from reactive tasks enables appropriate human intervention in autonomous operations. At each tier, mechanisms are provided for flexible response to novel events. We demonstrated the effectiveness of 3T to control life support systems for NASA's Lunar/Mars Life Support Test Project View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An intelligent landing controller for a simulated running jointed leg

    Page(s): 226 - 231
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (128 KB)  

    Controlling the running movement of a machine with jointed legs is a difficult problem in robotics. The control of a single running stride can be broken down into three phases: takeoff, ballistic, and landing. The paper describes the development of an intelligent controller for the landing phase of a simulated jointed leg's running stride. The landing controller takes control when the foot touches the ground after the airborne ballistic phase of the stride. It learns from experience to control the leg so as to recover from the impact with the ground and to reposition the leg for the takeoff of the next running stride, Very accurate control is achieved even during the first attempt at the running stride View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Evolution of neural networks for the detection of breast cancer

    Page(s): 34 - 40
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (124 KB)  

    This paper is based on a modified form of Fogel's evolutionary programming approach for evolving neural networks for the detection of breast cancer using fine needle aspirate data. A data visualization and preprocessing description is given, which not only depicts the benign and malignant raw data in graphical interpretative form but also includes a “symmetrized dot pattern” of this same data which may be used to corroborate the classification provided by the network. These evolved architectures routinely achieved a greater than 96% classification accuracy while, at the same time, achieving a much smaller type II error (calling a malignant sample benign). These results were obtained with different data sets using the same architecture, and were also obtained with the same data set over a family of evolved architectures View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Natural language sensing for autonomous agents

    Page(s): 374 - 381
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (100 KB)  

    In sufficiently narrow domains, natural language understanding may be achieved via an analysis of surface features without the use of a traditional symbolic parser. We illustrate this notion by describing the natural language sensing in Virtual Mattie (VMattie), an autonomous clerical agent. VMattie “lives” in an UNIX system, communicates with humans via e-mail in natural language with no agreed upon protocol, and autonomously carries out her tasks without human intervention. VMattie's limited domain allows for surface level natural language processing. VMattie's language understanding module has been implemented as a copycat-like architecture though her understanding takes place differently. The mechanism includes a slipnet storing domain knowledge and a pool of codelets templates for building and verifying understanding. We describe the design and implementation of natural language sensing for autonomous agents such as VMattie View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Coal molecular structure construction by genetic algorithm

    Page(s): 111 - 115
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (92 KB)  

    It has been clarified that the coal molecular consists of the following three components: the aromatic class cluster, the bridge connecting the aromatic class cluster, and the end permutation radical. In this paper, we propose a method of constructing the molecular structure using the genetic algorithm (GA). In this method, the chromosomes are expressed in the form of graphs. At the crossover operation, partial graph of one parent is kept and the nodes in the rest graph inherit connectivity of nodes from the other parent. The evaluation function is determined based on the heuristics of chemists as “the density of the coal molecular is averagely uniform”. From the result of experiments carried out with structures constructed from real coal data, it is shown that our method is effective to estimate the coal molecular structure View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Application of symmetry patterns to recognition of functional sites and phylogenetic analysis of DNA sequences

    Page(s): 152 - 157
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (688 KB)  

    The set of genetic texts, taking part in replication initiation has been analyzed from the point of view of their symmetry. The subsets, corresponding to the phenotypical classification, were found to contain the same symmetrical structures. Along with the traditional symmetries (potential hairpins, direct and inverted repeats) the purine-pyrimidine and aminoketo repeats were found at equal distances in related sequences. Among the symmetrical structures the replication protein binding sites have been identified as well as sites taking part in the transcription initiation. The results obtained show the efficiency of symmetrical analysis in solving the problem of pattern recognition in genetic text View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Robust speech recognition using a noise rejection approach

    Page(s): 326 - 335
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (32 KB)  

    In this paper, we explore some new approaches to improve speech recognition accuracy in a noisy environment. The key approaches taken are: (a) use no additional data (i.e. use only speakers data, no data for noise) for training and (b) no adaptation phase for noise. Instead of making adaptation in the recognition, preprocessing or both stages, we make a noise tolerant (rejection) speech recognition system where the system tries to reject noise automatically because of its inherent structure. We call our approach a noise rejection-based approach. Noise rejection is achieved by using multiple views and dynamic features of the input sequences. Multiple views exploit more information from the available data that is used for training multiple HMMs (hidden Markov models). This makes the training process simpler, faster and avoids the need to use a noise database, which is often difficult to obtain. The dynamic features (added to the HMM using vector emission probabilities) add more information about the input speech during training. Since the values of dynamic features of noise are usually much smaller than that of the speech signal, it helps reject the noise during recognition. Also, multiple views of the input sequence are applied to multiple HMMs during recognition and the outcome of the multiple HMMs are combined using maximum evidence criterion. Our tests show very encouraging results. We also incorporate higher level decision making to more judiciously combine the outcomes of the multiple HMMs to further improve the accuracy. For this, we use meta reasoning to identify the problem complexity and accordingly allocate resources View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural-network-based compression algorithm for gray scale images

    Page(s): 422 - 428
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (152 KB)  

    This paper presents an image compression algorithm for gray scale images, based on neural networks. According to this algorithm the image will be first decomposed into Hadamard set of functions and second, the coefficients from the decomposition will be dynamically clustered by a newly proposed dynamic adaptive clustering method (DACM). We show that DACM converges to approximate the optimum solution based on the least sum of squares criterion theoretically and experimentally. We applied the compression method to various gray scale images and show its efficiency in providing high compression rates. In order to show some comparative results for the proposed method, we have chosen the well-known JPEG. Its algorithm has similar structure and therefore is a good basis for comparison. The results from the gray scale images experiments are in favor of the proposed method View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Formal model for real time diagnosis of dynamic systems

    Page(s): 218 - 225
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1376 KB)  

    The task of real-time causal diagnosis of disturbances has been conceived traditionally from a procedural point of view, in the sense that the attention is focused on developing efficient procedures capable of evaluating the state of some variables of the system so that real time objectives imposed were satisfied. The main handicap of these methods lies in the difficulty to plan the diagnostic process, particularly when a high number of variables are to be observed model-based diagnosis constitutes a more complete approach to the topic. Considering the availability of a simulated model of the system, the task of the diagnostic procedure is now performed on the simulated model, facilitating the observation and handling of the variables of the system. However, the absence of languages allowing us to develop simulated models of real systems limits the use of this theory to simple cases. An approach to real-time causal diagnosis of dynamic systems based on a pre-established planning of any possible diagnostic situation in such a way real-time objectives are satisfied, is presented in this work. Artificial intelligence techniques, particularly inductive methods have been considered according to two essential steps: formulation of the causal diagnostic model, specifying the particular characteristics of the problem in hand; and generation of an information structure according to the characteristics of the formulated model, whose performance will guarantee the diagnostic objectives View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Gene family identification network design

    Page(s): 103 - 110
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (244 KB)  

    The exponential accumulation of molecular data will facilitate the discovery of new knowledge by using information embedded within families of homologous sequences. As an approach to the management and analysis of sequence data, we have developed an integrated system, termed GeneFIND (Gene Family Identification Network Design), for database searching against gene families. It provides rapid and accurate protein family identification by combining global and motif sequence similarities and incorporating ProClass family information. Multilevel filters are used, starting with the MOTIFIND neural network and BLAST search, followed by SSEARCH alignment motif pattern match, hidden Markov modeling of motifs and ClustalW motif alignment. GeneFIND has been implemented as a full-scale system for the classification of more than 1000 ProSite and 3000 PIR families. It is used to identify thousands of new family members and is well suited for genomic sequence analysis View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Crossover operators with adaptive probability

    Page(s): 10 - 16
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (4 KB)  

    Genetic algorithms (GAs) are adaptive methods, which can be employed to solve search and optimization problems. The GA relies on genetic operators to exchange gene between individuals for generating better offspring. An important issue to execute GA efficiently is to maintain population diversity and to sustain local improvement in the search stage. However, both effects always hinder each other. We propose to apply different kinds of crossover operators, i.e. arithmetic and BLX-α crossovers, to control the diversity and convergence of the GA in continuous-space framework. We also utilize self-adaptation method to control the probability of crossover such that the balance of exploitation and exploration can be kept. It is shown empirically that the proposed methods outperform the classical GA strategy on several benchmark functions View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Error repair in human handwriting: an intelligent user interface for automatic online handwriting recognition

    Page(s): 389 - 395
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (124 KB)  

    Since both users and recognition algorithms make mistakes, it is desirable for the user interface of a handwriting recognition system to have mechanisms recovering from errors. We address the problem of error repair in online handwriting recognition. First, we perform a user study on error occurrences and corresponding repair patterns in human handwriting. Based on a data analysis, we have identified typical types of errors and repair patterns. We then propose methods to deal with error repair in an online handwriting system. We have developed a prototype system to demonstrate and evaluate the proposed error handling mechanisms. The system extends NPen++, an online handwriting recognition system developed in our lab, by providing error repair abilities to users in addition to its high recognition rate. The experimental results indicate that the error handling mechanisms can significantly improve the system performance in case of the data containing error repair View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Biologically-based learning in the ARBIB autonomous robot

    Page(s): 49 - 56
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (56 KB)  

    We describe the autonomous robot ARBIB, which uses biologically-motivated forms of learning to adapt to its environment. The “nervous system” of ARBIB has a nonhomogeneous population of spiking neurons, and uses both nonassociative (habituation, sensitization) and associative (classical conditioning) forms of learning to modify pre-existing (“hard-wired”) reflexes. As a result of interaction with its environment, interesting and “intelligent” light-seeking and collision-avoidance behaviors emerge which were not pre-programmed into the robot-or “animat”. These behaviors are similar to those described by other workers who have generally used behaviorally-motivated reinforcement learning rather than biologically-based associative learning. The complexity of observed behavior is remarkable given the extreme simplicity of ARBIB's “nervous system”, having just 33 neurons. It does not even have a brain! We take this to indicate that great potential exists to explore further “the animat path to AI” View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Implementation of a multi-fingered robotic hand using a method of fuzzy blocks

    Page(s): 262 - 267
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (956 KB)  

    The control of a mechanical robotic hand is plagued by an inability to derive accurate dynamic models of such a mechanism. Static and kinetic friction are major obstacles in the control of a mechanical hand. The paper presents a fuzzy based controller which has the ability to automatically regenerate the member set during the translation of an arbitrary joint in a robotic hand. This control system has been implemented on an IBM compatible computer using a custom designed acquisition/conversion interface and a mechanical hand assembly. A series of experiments have been conducted which verify that the method of fuzzy blocks was successful for controlling joint position. The simulation results include bath fine movements, needed for dexterity, and gross movements which can be used for grasping. This system provides excellent results and it is extremely cost effective View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The dynamic predictive memory architecture: integrating language with task execution

    Page(s): 202 - 206
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (104 KB)  

    The dynamic predictive memory architecture (DPMA) is an architecture for integrating natural language understanding with dynamic task execution. We describe the architecture and its components, as well as issues which became evident while implementing a control station for a free-flying camera View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Connecting perception and action by associating symmetries in vision and language

    Page(s): 309 - 314
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (28 KB)  

    Symmetries are of fundamental importance in information processing. It is well understood how important visual symmetries are in the understanding of scenes. This paper explores the symmetries of languages of actions, and relates them to visual symmetries. Linguistic symmetry is explored in the context of problem solving. Algebraic techniques permit the representation and manipulation of linguistic symmetry, and the exploration of its connection to perceptual symmetry. It is shown that the invariants of the decomposition of a language of actions form a complete and independent set of perceptual features that naturally describe the search space View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Method to achieve better performance in genetic algorithms applied to time-constrained, multi-solution problems

    Page(s): 2 - 9
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1100 KB)  

    It has been demonstrated that a periodic complete reinitialization of a running genetic algorithm (GA) solution will result in a higher convergence rate for a series of problems. This technique, referred to as the “total-comet-strike” operator, has been applied to a number of multi-solution GA problems. Where it has been used, an improvement has been shown both in the number of cases that converge within an imposed time limit and in the average time required for each case View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.