Fundamenta Informaticae - Volume 72, issue 4

Authors: Collings, Simon | Kozera, Ryszard | Noakes, Lyle

Article Type: Research Article

Abstract: This paper shows that a strictly convex solid under observation by two or more cameras with known extrinsic and intrinsic parameters contains points on its surface calculable in terms of multiple photographic images. In the case of a solid with a C^1 boundary we can also extract surface orientation at these points, and when the solid has a C^2 boundary and a known reflectance map, curvature information can also be calculated.

Keywords: N-view shape recovery, convex solid, visual hull, occluding edge

Citation: Fundamenta Informaticae, vol. 72, no. 4, pp. 437-452, 2006

Authors: Fredriksson, Kimmo

Article Type: Research Article

Abstract: We consider approximate pattern matching in natural language text. We use the words of the text as the alphabet, instead of the characters as in traditional string matching approaches. Hence our pattern consists of a sequence of words. From the algorithmic point of view this has several advantages: (i) the number of words is much less than the number of characters, which in effect means shorter text (less possible matching positions); (ii) the pattern effectively becomes …shorter, so bit-parallel techniques become more applicable; (iii) the alphabet size becomes much larger, so the probability that two symbols (in this case, words) match is reduced. We extend several known approximate string matching algorithms for this scenario, allowing k insertions, deletions or substitutions of symbols (natural language words). We further extend the algorithms to allow k' errors inside the pattern symbols (words) as well. The two error thresholds k and k' can be applied simultaneously and independently. Hence we have in effect two alphabets, and perform approximate matching in both levels. From the application point of view the advantage is that the method is flexible, allowing simple solutions to problems that are hard to solve with traditional approaches. Finally, we extend the algorithms to handle multiple patterns at the same time. Depending on the search parameters, we obtain algorithms that run in linear or sublinear time and that perform the optimal number of word comparisons on average, We conclude with experimental results showing that the methods work well in practice. Show more

Keywords: Approximate string matching, filtering, natural language processing, phrase matching, word alphabet

Citation: Fundamenta Informaticae, vol. 72, no. 4, pp. 453-466, 2006

Authors: Păun, Andrei | Popa, Bianca

Article Type: Research Article

Abstract: This work is a continuation of the investigations aiming to bridge membrane computing (where in a compartmental cell-like structure the chemicals to evolve are placed in the compartments defined by the membranes) and brane calculi (where one considers again a compartmental cell-like structure with the chemicals/proteins placed on the membranes themselves). In the current paper we use objects both in compartments and on membranes (the latter are called proteins), with the objects from membranes evolving under …the control of the proteins. Several possibilities are considered (objects only moved across membranes or also changed during this operation, with the proteins only assisting the move/change or also changing themselves). Somewhat expected, computational universality is obtained for several combinations of such possibilities. Show more

Keywords: Membrane computing, brane calculus, Turing computability, universality

Citation: Fundamenta Informaticae, vol. 72, no. 4, pp. 467-483, 2006

Authors: Potikas, Petros | Rondogiannis, Panos | Gergatsoulis, Manolis

Article Type: Research Article

Abstract: The branching-time transformation is a recent technique for optimizing Chain Datalog programs. In this paper we propose a significant extension of the branching-time transformation which we believe opens up a promising new direction of research in the area of value-propagating Datalog optimizations. More specifically, the proposed transformation can handle more general programs that allow multiple consumptive occurrences of variables in the bodies of clauses. This extension is achieved by using as target language …the temporal logic programming formalism Datalog_{nS} enriched with choice predicates (a non-deterministic construct that was originally introduced in the area of intensional logic programming). We demonstrate the correctness of the transformation and propose several optimizations that can be applied to the target code. Moreover, we define a bottom-up proof procedure that applies to the target programs and demonstrate that it always terminates (despite the fact that the Herbrand base of these programs is generally infinite). Show more

Keywords: Logic Programming, Deductive Databases, Program and Query Transformations, Non-deterministic Constructs

Citation: Fundamenta Informaticae, vol. 72, no. 4, pp. 485-527, 2006

Authors: Shen, Haifeng | Li, Qunxia | Guo, Jun | Liu, Gang

Article Type: Research Article

Abstract: This paper proposes a novel robust speech recognition approach based on the model-based feature compensation. The approach combines the GMM-based feature compensation and the HMM-based feature compensation together and employs the multiple recognition passes to achieve the best performance. In the initial recognition procedure, the GMM-based feature compensation approach is employed to give better clean model and noise model. Then we further refine these models by employing the HMM-based feature compensation approach. The …statistical model of the clean speech and the noise is combined by using vector Taylor series (VTS) approximation. The experimental results show that the novel approach makes a significant improvement compared to the GMM-based feature compensation and the HMM-based feature compensation without any compensation in the initial pass. Show more

Keywords: robust speech recognition, feature compensation, EM algorithm

Citation: Fundamenta Informaticae, vol. 72, no. 4, pp. 529-539, 2006