December 2000, Volume 9, Number 4
The paper deals with the problem of automatic phonetic segmentation of
speech signals, namely for speech analysis and recognition purposes.
Several methods and approaches are described and evaluated from the
point of view of their accuracy. A complete instruction for creating an
annotated database for training a Czech speech recognition system is
provided together with the authors' own experience. The results of the
work have found practical applications, for example, in developing a
tool for semi-automatic speech segmentation, building a
large-vocabulary phoneme-based speech recognition system and designing
an aid for learning and practicing pronunciation of words or phrases in
the native or a foreign language.
Four modulation schemes, namely minimum shift keying (MSK), Gaussian minimum shift keying (GMSK), multiamplitude minimum shift keying (MAMSK) and π/4 differential quadrature phase shift keying (π/4-QPSK) are described and their applicability to wireless access is discussed in the paper. Low complexity receiver structures based on differential detection are analysed to estimate the performance of the modulation schemes in the additive Gaussian noise and the Rayleigh and Rice envelope fast fading channel. The bandwidth efficiency is calculated to evaluate the modulation schemes. The results show that the MAMSK scheme gives the greatest bandwidth efficiency, but its performance in the Rayleigh channel is rather poor. In contrast, the MSK scheme is less bandwidth efficient, but it is more resistant to Rayleigh fading. The performance of π/4-QPSK signal is considerably improved by appropriate prefiltering.
In this paper, several adaptive least mean squares (LMS) location-invariant filter (L-filter) modifications will be described. These filters are based on linear combination of order statistics. The adaptive L-filters are able to adapt well to variety of noise probability distribution, including impulsive noise. They also perform well in the case of nonstationary signals and, therefore, they are suitable for image processing, too. Following this L-filter property, applications of the adaptive LMS L-filters for filtering two-dimensional static images degraded by mixed noise consisting of additive Gaussian white noise and impulsive noise will be presented in this paper. Based on conveniently selected experiments intent on image filtering, the properties of a several adaptive L-filters modifications will be demonstrated and compared. It will follow from experiment results, that the L-filter modification called signal-dependent LMS L-filter yields the best results.
In this paper are presented some results of implementation of digital watermarking methods into image coding based on fractal principles. The paper focuses on two possible approaches of embedding digital watermarks into fractal code of images - embedding digital watermarks into parameters for position of similar blocks and coefficients of block similarity. Both algorithms were analyzed and verified on gray scale static images.
Block motion estimation using full search algorithm is computationally extensive. Previously proposed fast algorithms reduce the computation cost by limiting the number of locations searched. In this paper we present algorithms for block motion estimation that produce similar performance to that full search algorithm. The algorithms are based on the pixel decimation.
This paper presents an investigation of the deterministic and stochastic chaos. The modified Colpitts oscillator is used as an example of deterministic chaos in electronic circuits. The graphical method of Lorenz maps is used for graphical observation of both chaotic classes.
State models of multiple-input linear non-autonomous dynamical systems
are modified to the autonomous form using a multiple piecewise-linear
(PWL) feedback. By such a way the corresponding state models of PWL
autonomous dynamical system topologically conjugate to Lur'e systems
 can be derived. As an example, the canonical state models of
multiple-input linear and the corresponding multiple PWL feedback
systems are proposed. Their state matrix equations and the
integrator-based circuit models are shown.