Showing papers by "Helsinki University of Technology published in 1988"
TL;DR: In this paper, the authors describe a self-organizing system in which the signal representations are automatically mapped onto a set of output responses in such a way that the responses acquire the same topological order as that of the primary events.
Abstract: This work contains a theoretical study and computer simulations of a new self-organizing process. The principal discovery is that in a simple network of adaptive physical elements which receives signals from a primary event space, the signal representations are automatically mapped onto a set of output responses in such a way that the responses acquire the same topological order as that of the primary events. In other words, a principle has been discovered which facilitates the automatic formation of topologically correct maps of features of observable events. The basic self-organizing system is a one- or two-dimensional array of processing units resembling a network of threshold-logic units, and characterized by short-range lateral feedback between neighbouring units. Several types of computer simulations are used to demonstrate the ordering process as well as the conditions under which it fails.
8,247 citations
TL;DR: A brief survey of the motivations, fundamentals, and applications of artificial neural networks, as well as some detailed analytical expressions for their theory.
1,418 citations
TL;DR: A speaker-adaptive system that transcribes dictation using an unlimited vocabulary is presented that is based on a neural network processor for the recognition of phonetic units of speech.
Abstract: The factors that make speech recognition difficult are examined, and the potential of neural computers for this purpose is discussed. A speaker-adaptive system that transcribes dictation using an unlimited vocabulary is presented that is based on a neural network processor for the recognition of phonetic units of speech. The acoustic preprocessing, vector quantization, neural network model, and shortcut learning algorithm used are described. The utilization of phonotopic maps and of postprocessing in symbolic forms are discussed. Hardware implementations and performance of the neural networks are considered. >
647 citations
24 Jul 1988
TL;DR: Three basic types of neural-like networks, backpropagation network, Boltzmann machine, and learning vector quantization, were applied to two representative artificial statistical pattern recognition tasks, each with varying dimensionality.
Abstract: Three basic types of neural-like networks (backpropagation network, Boltzmann machine, and learning vector quantization), were applied to two representative artificial statistical pattern recognition tasks, each with varying dimensionality. The performance of each network's approach to solving the tasks was evaluated and compared, both to the performance of the other two networks and to the theoretical limit. The learning vector quantization was further benchmarked against the parametric Bayes classifier and the k-nearest-neighbor classifier using natural speech data. A novel learning vector quantization classifier called LVQ2 is introduced. >
496 citations
TL;DR: A tutorial review is given of some of the basic electronic properties of two-dimensional systems and their physical realisations.
Abstract: A tutorial review is given of some of the basic electronic properties of two-dimensional systems and their physical realisations.
370 citations
363 citations
TL;DR: In this article, a new mixed finite element formulation for the equations of linear elasticity is considered, where the variables approximated are the displacement, the unsymmetric stress tensor and the rotation.
Abstract: A new mixed finite element formulation for the equations of linear elasticity is considered. In the formulation the variables approximated are the displacement, the unsymmetric stress tensor and the rotation. The rotation act as a Lagrange multiplier introduced in order to enforce the symmetry of the stress tensor. Based on this formulation a new family of both two-and three-dimensional mixed methods is defined. Optimal error estimates, which are valid uniformly with respect to the Poisson ratio, are derived. Finally, a new postprocessing scheme for improving the displacement is introduced and analyzed.
286 citations
TL;DR: Spatial resolution of magnetoencephalography was studied by computer simulations using a spherical conductor model for the head to illustrate the dependence of spatial resolution on several factors including noise, source depth, source strength, flux transformer configuration and the choice of the measurement locations.
182 citations
TL;DR: In this article, the atomic layer epitaxy method was used to grow cubic NbNbN and cubic TaN at 773 K from the corresponding chlorides and ammonia by using the nuclear resonance broadening technique and compared with structure and other properties.
174 citations
TL;DR: To increase the shape control in the longitudinal surface direction, projection curve techniques for interactive B-spline surface interpolation are presented.
Abstract: The skinning method for interpolation over user-defined surface sections is one of the most widely used tools for interactive shape description in computer-aided design. The standard interaction technique with the skinning method uses a spine curve for the transformation of planar section curves to the object space. To increase the shape control in the longitudinal surface direction, projection curve techniques for interactive B-spline surface interpolation are presented. The methods are demonstrated with an application from ceramic industry.
134 citations
TL;DR: In this paper, Gaussian Schell-model sources were analyzed with an extended ray-transfermatrix method and algebraic expressions were derived for the location, size, and coherence area of the image waist and for the depth of focus and the far-field diffraction angle.
Abstract: Imaging of Gaussian Schell-model sources by general lossless systems is analyzed with an extended ray-transfermatrix method. Algebraic expressions are derived for the location, size, and coherence area of the image waist and for the depth of focus and the far-field diffraction angle. These results are shown to provide a continuous transformation between laser-beam optics and geometrical optics. They also lead naturally to several equivalence and invariance relations pertaining to isotropic and anisotropic Gaussian Schell-model sources. As an application, the importance of effects due to partial spatial coherence in beam focusing is examined.
TL;DR: In this article, the authors address the problem of optimal selection of shape functions for p-type finite elements and discuss the effectivity of the conjugate gradient and multilevel iteration method for solving the corresponding linear system.
Abstract: : The paper addresses the question of the optimal selection of the shape functions for p-type finite elements and discusses the effectivity of the conjugate gradient and multilevel iteration method for solving the corresponding linear system. The selection of the shape functions is of major importance for the performance of the solver based on iterative methods. Neither the theory nor practice of the optimal selection of the shape functions is available yet. We have seen that the condensation approach which has obvious advantages from the point of parallel computations is a very effective tool for keeping the condition number under the control and is especially advantageous for the conjugate gradient method.
01 Jan 1988
TL;DR: A property which is commonplace in the brain but which has always been ignored in the “learning machines” is a meaningful order of their processing units.
Abstract: A property which is commonplace in the brain but which has always been ignored in the “learning machines” is a meaningful order of their processing units. “Ordering” thereby usually does not mean moving of units to new places. The units may even be structurally identical; the specialized role is determined by their internal parameters which are made to change in certain processes. It then appears as if specific units having a meaningful organization were produced.
TL;DR: The small interaction between N 100m and N100m' and their different recovery cycles suggest that different activation patterns underlie these two 100 msec responses at the auditory cortex.
TL;DR: The results suggest that the masking effects of continuous noise, seen at the auditory cortex, derive mainly from the periphery whereas the effects of sounds with intensity and frequency modulations take place at more central auditory pathways.
Abstract: We have studied the effects of masking sounds on auditory evoked magnetic fields (AEFs) of healthy humans. The AEFs were elicited by 25-ms tones presented randomly to the left or to the right ear, and the responses were recorded over the right auditory cortex. Without masking, the 100-ms deflection (N100m) was of somewhat higher amplitude and of shorter latency for contra- than ipsilateral stimuli. Continuous speech, music, or intermittent noise, delivered to the left ear, dampened N100m to stimulation of both ears without correlated changes in sensation. Intermittent noise had a weaker effect on N100m than speech or music. Continuous noise fed to the left ear dampened both the sensation of and the responses to the left-ear stimuli, with no significant effect on the responses to the right-ear stimuli. The results suggest that the masking effects of continuous noise, seen at the auditory cortex, derive mainly from the periphery whereas the effects of sounds with intensity and frequency modulations take place at more central auditory pathways.
TL;DR: Lignin peroxidase was produced by nylon-web immobilized Phanerochaete chrysosporium in carbon-limited medium in a modified Biostat®E bioreactor using Veratryl alcohol as an enzyme production activator and some homology was observed between P. chrysOSporium and Pleurotus ostreatus ligninolytic enzymes.
TL;DR: It is shown that the positron trapping and annihilation in the native defects are strongly dependent on the position of the Fermi level in as-grown n-type GaAs.
Abstract: Native vacancy defects are studied in as-grown GaAs materials by positron-lifetime measurements. Direct evidence of native monovacancy-type defects is found in as-grown n-type Te-, Sn-, and Si-doped GaAs. However, the same evidence is not found in as-grown semi-insulating In- or Cr-doped GaAs or semi-insulating undoped GaAs or p-type GaAs(Zn). It is shown that the positron trapping and annihilation in the native defects are strongly dependent on the position of the Fermi level in as-grown n-type GaAs. It is concluded that the configurations of the native monovacancy defects in GaAs (Te or Sn) change with the position of Fermi level. Two Fermi-level-controlled transitions are found: one is located at 0.035\ifmmode\pm\else\textpm\fi{}0.015 eV and the other at 0.10\ifmmode\pm\else\textpm\fi{}0.02 eV below the conduction band. It is proposed that the two transitions correspond to two charge-state transitions of the arsenic vacancy ${V}_{\mathrm{As}{}^{2\mathrm{\ensuremath{-}}}\ensuremath{\rightarrow}{V}_{\mathrm{As}{}^{\mathrm{\ensuremath{-}}}}}$ and ${V}_{\mathrm{As}{}^{\mathrm{\ensuremath{-}}}\ensuremath{\rightarrow}{V}_{\mathrm{As}{}^{0}}}$, respectively.
TL;DR: In this paper, a system for the accurate measurement of high-frequency periodic time-domain voltage and current waveforms of a nonlinear microwave device is presented, where measurements are performed in the time domain using a high-speed sampling oscilloscope.
Abstract: A system is presented for the accurate measurement of high-frequency periodic time-domain voltage and current waveforms of a nonlinear microwave device. The measurements are performed in the time domain using a high-speed sampling oscilloscope. The results are Fourier-transformed into the frequency domain for error correction and then back into the time domain. An error-correction algorithm is presented that makes it possible to obtain accurate waveforms in spite of nonideal system components. Practical difficulties in measurement-system characterization are also discussed. An accurate circuit model for the measurement fixture is developed and its element values are determined. Measurement results are given showing the waveforms in a microwave transistor operated in the nonlinear region. The errors caused by signal processing are discussed. >
TL;DR: The possible planar superfluid B-B boundaries between inequivalent B-phase vacua are considered and diabolical points of codimension 4 are discussed, characterized by the occurrence of a contact between the different branches of the quasiparticle spectra; in the present case, the branches of particles and holes.
Abstract: The possible planar superfluid B-B boundaries between inequivalent B-phase vacua are considered; such B-B interfaces provide an analogy with the cosmic domain walls that are believed to have precipitated in the phase transitions of the early Universe. Several of them display nontrivial structure in (k,r) space (i.e., the union of the momentum and real spaces). Such a wall represents an instanton connecting two B-phase vacua with different k-space topology. The transition between the vacua occurs through the formation of a pointlike defect either in the (k,r) space, or in the (k,t) space. These defects are so-called diabolical points of codimension 4, at which the fermionic energy tends to zero, thus providing the fermionic zero modes. Such points are new examples (within condensed-matter physics) of the peculiar diabolical points, which are characterized by the occurrence of a contact between the different branches of the quasiparticle spectra; in the present case, the branches of particles and holes, respectively. These points are here discussed for the case of the superfluid phases of liquid $^{3}\mathrm{He}$ in close analogy with the quantum field theory of fermions interacting with classical bosonic fields. The cosmiclike domain walls in superfluid $^{3}\mathit{B}$ are observable in principle; in particular, the motion of the superfluid A-B interface is governed at low temperatures by the periodical emission of these topological excitation planes. The edges of B-B interfaces serve to generate fractionally quantized pure and mixed mass and spin supercurrent vortices in $^{3}\mathit{B}$, while holes in these surfaces may give rise to the corresponding vortex rings and combined vortex and/or spin-disclination rings.
TL;DR: The widely used SAMPO Ge(Li) and HPGe gamma spectrum analysis program has been adapted to IBM Personal Computers in a thoroughly revised version MicroSAMPO, intended for peak search, peak fitting, nuclide identification and activity calculations.
Abstract: The widely used SAMPO Ge(Li) and HPGe gamma spectrum analysis program has been adapted to IBM Personal Computers in a thoroughly revised version MicroSAMPO. The program is intended for peak search, peak fitting, nuclide identification and activity calculations. The use of calibrated peak shape functions for peak area determination makes it possible to resolve complex multiplets with strongly overlapping peaks. Menu-driven user interface, colour graphics diplays, semi-automatic calibrations and interactive options have been designed to make the program more user-friendly. It is well suited for both spectroscopic research and routine analysis.
TL;DR: In this paper, a mesh dependent energy norm was used for analyzing mixed finite element methods for the plate problem, which applies both to the Kirchhoff and to the Mindlin-Reissner formulation of the problem.
Abstract: We set up a framework for analyzing mixed finite element methods for the plate problem using a mesh dependent energy norm which applies both to the Kirchhoff and to the Mindlin-Reissner formulation of the problem. The analysis techniques are applied to some low order finite element schemes where three degrees of freedom are associated to each vertex of a triangulation of the domain. The schemes proceed from the Mindlin-Reissner formulation with modified shear energy.
TL;DR: The distributions of H atoms, displaced Si atoms, and vacancy-type defects in Si(100) produced by the implantation, at room temperature, of 1-2-3-4-5-6-7 were measured with use of ion and slow-positron beam techniques.
Abstract: The distributions of H atoms, displaced Si atoms, and vacancy-type defects in Si(100) produced by the implantation, at room temperature, of 1\ifmmode\times\else\texttimes\fi{}${10}^{16}$ 35-, 60-, and 100-keV ${\mathrm{H}}^{+}$-ions/${\mathrm{cm}}^{2}$ were measured with use of ion and slow-positron beam techniques. Three different regions of damage were observed. The damage distribution did not correlate with the deposited energy distribution. A vacancy-type damage region surviving room-temperature annealing and the dynamic annealing during implantation is produced in the region where the energy of the primary Si recoils exceeds a threshold value of 2.0\ifmmode\pm\else\textpm\fi{}0.5 keV, i.e., an energy spike is needed for damage formation. In the region of the deposited energy peak, H is associated with damage consisting of SiH centers and vacancy complexes. At the end region of the implantation range, the distribution of displaced Si atoms is produced by H impurities in crystalline Si. Two different recovery stages were observed. The first stage, at 500 K, is associated with the damage region containing vacancy-type defects and SiH centers with vacancy complexes; the recovery energy is 1.7 eV. The second stage, at 700 K, is attributed to the end region of the H range, where the recovery energy associated with the recrystallization of Si after the H loss is 2.1 eV.
TL;DR: In this article, the effects of partial spatial coherence on the propagation and focusing of gaussian Schell-model beams were investigated in free space and optical optical systems, and the measured coherence-dependences of the beam profile and the focal shift and image spot size were shown to be in good agreement with recent theoretical predictions.
TL;DR: In this paper, a photolithographic technique for fabrication of these elements was also presented, and its accuracy was analyzed with the aid of numerical simulations and actual test results, with the help of simulated annealing and damped least squares algorithms.
Abstract: Optimisation of certain holographic 1 to N and 1 to N*N beamsplitters (Dammann gratings) with the aid of simulated annealing and damped least-squares algorithms is considered. Structures with N=5 to 53 uniform beams have been calculated. A photolithographic technique for fabrication of these elements is also presented, and its accuracy analysed with the aid of numerical simulations and actual test results.
TL;DR: In this paper, the design of a complete dc SQUID with a flux transformer input circuit is discussed, which allows simultaneous optimization of the SQUID parameters, minimizing the parasitic capacitance, control over the resonances, and good inductance matching to practical magnetometer coils.
Abstract: The design of a complete dc SQUID with a flux transformer input circuit is discussed. The flux coupling circuits introduce a substantial capacitance across the SQUID and give rise to many resonances which may couple strongly to the SQUID dynamics. Both effects lead to multiple modes in the SQUID dynamics and consequently to excess noise. For a low-noise SQUID with smooth characteristics, our analysis and practical considerations suggest signal coupling via an intermediary transformer. This method allows simultaneous optimization of the SQUID parameters, minimizing the parasitic capacitance, control over the resonances, and good inductance matching to practical magnetometer coils. A model is developed to optimize the structure: it describes the whole circuit with the help of a suitably modified autonomous SQUID, provided that the system is free from multiple modes due to resonances or large parasitic capacitance. Following these design principles, we have built a dc SQUID, primarily for use in biomagnetic research, but also well suited for other applications. The fabrication of the SQUID and the high-quality electronics especially suitable for multiple-SQUID devices is presented. The SQUIDs showed smooth characteristics, and the lowest measured noise of our complete SQUID is % MathType!MTEF!2!1!+-% feaafeart1ev1aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLn% hiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr% 4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq-Jc9% vqaqpepm0xbba9pwe9Q8fs0-yqaqpepae9pg0FirpepeKkFr0xfr-x% fr-xb9adbaqaaeGaciGaaiaabeqaamaabaabaaGcbaGaaGymaiaac6% cacaaIZaGaey41aqRaaGymaiaaicdadaahaaWcbeqaaiabgkHiTiaa% iAdaaaGccqqHMoGrdaWgaaWcbaGaaGimaiaac+caaeqaaOWaaOaaae% aacaWGibGaamOEaaWcbeaaaaa!428B!$1.3 \times 10^{ - 6} \Phi _{0/} \sqrt {Hz} $, indicating the success of the design.
TL;DR: In this article, the authors consider time-dependent perturbations of generators of strongly continuous semigroups on a Banach space and show that the solutions of a generalised variation-of-constants formuladefine an evolutionary system.
Abstract: We consider time-dependent perturbations of generators of strongly continuous semigroups on a Banach space. The perturbations map the Banach space into a bigger space, which is the second dual of the original space in a specific semigroup sense. Using the theory of dual semigroups we show that the solutions of a generalised variation-of-constants formuladefine an evolutionary system. We investigate continuity and differentiability propertiesof this evolutionary system and its dual system and examine in what sense the perturbed generator and its adjoint generate these evolutionary systems. It is shown that the results apply naturally to retarded functional differential equations and age structured population dynamics.
11 Apr 1988
TL;DR: A microprocessor-based real-time speech recognition system that is able to produce orthographic transcriptions for arbitrary words or phrases uttered in Finnish or Japanese and can also be used as a large-vocabulary isolated word recognizer.
Abstract: A microprocessor-based real-time speech recognition system is described. It is able to produce orthographic transcriptions for arbitrary words or phrases uttered in Finnish or Japanese. It can also be used as a large-vocabulary isolated word recognizer. The acoustic processor of the system transcribing speech into phonemes is based on neural network principles. The so-called phonotopic maps constructed by a self-organizing process are employed. The coarticulation effects in phonetic transcriptions are compensated by means of automatically derived rules which describe the morphology of errors at the acoustic processor output. Without applying any language model, the recognition result is correct up to 92 or even 97 per cent referring to individual letters. >
TL;DR: In this paper, the microwave response to different surface and forest types was investigated using the Nimbus-7 Scanning Multichannel Microwave Radiometer (Nimbus7 SMMR).
Abstract: 10.7-, 18-, and 37-GHz data from the Nimbus-7 Scanning Multichannel Microwave Radiometer (Nimbus-7 SMMR) were used to investigate the microwave response to different surface and forest types. SMMR data for the fall seasons of 1978 through 1981 were compared against a digital surface type map that shows seven different surface types for southern Finland and six for northern Finland. For each land-cover category, the brightness temperature behavior as a function of frequency and polarization was determined. The precision of the results is evaluated, and they are compared to other published results. The discrimination of land-cover categories using the brightness temperature is discussed. >