Академический Документы
Профессиональный Документы
Культура Документы
ABSTRACT global basis functions are those whose value is not zero
on any subinterval. This class includes Walsh functions, numerical,
This article discusses the use of local systems of basic functions sawtooth;
defined on compact carriers. Their effectiveness is shown when localizable basis functions whose nonzero values are
solving the problems of discretization of finite signals. Signals can defined on nested segments. Examples are the HAAR and Hartmut
be not only time functions defined at finite intervals, but also functions.
functions of arguments of another physical nature, for example, The partition of the real axis is usually binary-rational. In
distances along surfaces. what follows, we shall mainly consider the interval [0, 1] or [0, 1]
and use the notion of a binary interval, which is obtained by
KEYWORDS dividing a given interval by 2рequal parts (P = 1, 2, ...):
Finite spectrum, finite function, signal energy, approximation,
HAAR wavelets, fast wavelet transform. j j 1
hk hpj p 1
2 2 p 1
INTRODUCTION (1)
p 1 p 1
The methods and techniques of digital signal processing raise the
interest of scientists and specialists working in various fields, such where,
j 0,1,...,2 , k j 2
as communications and control systems, radio engineering and Examples of binary segments are intervals
electronics, acoustics, geophysics and seismology, broadcasting [ 0;1]; [ 1/2; 3/4], [ 3/8; 4/8 ] and etc.
and television, measuring equipment and instrumentation. It also The length of the binary segment hpj is:
includes such relatively new directions in the creation of hardware
and software, such as audio and video signal processing, speech h , h
recognition, biometric systems, dynamic image processing, and where pj pj
are respectively its left and right halves
multimedia training systems. and also represent binary segments:
Traditional harmonic functions are widely used to build
j 1 2 j 1
2 j 1 j
the models of signals received from real objects. This is because hpj p 1 ; p
, hpj = p ;
many signals received from real objects can be easily represented 2 2 2 2 p 1
by a combination of sinusoidal and cosine waves, for which the (2)
Fourier analysis apparatus is used. The result is a transition from The system of Haar non-standardized functions is defined [2]:
time to frequency functions. However, the representation of the
time function by sinusoidal and cosine functions is only one of
many representations. Any complete system of orthogonal
functions can be used to expand into series that correspond to
Fourier series.
Widespread distributions in technical applications have
obtained orthogonal systems of continual disjointed basis functions
defined on the real axis, for which there are also algorithms of fast
transformations. They can be divided into two classes:
WOODSTOCK’18, June, 2018, El Paso, Texas USA H. Zaynidinov, M. Zaynutdinova, E. Nazirova.
har0 x 1
organizing minimum signal samples ensuring the necessary
accuracy of recovery, it is necessary to investigate the intrinsic
Important to note that spectra of the wavelet coefficients. Such systems of wavelet
The number P is called the order of the Haar functions functions as derivatives of the Gaussian function, Morlet wavelets,
It is known that the Haar series [2]: Shannon wavelets, etc. are theoretically determined on the whole
axis, but can be regarded as local. But the main role in the
f x C hark x
algorithms of discrete wavelet transform (DWT) is played by
k orthonormal wavelet bases based on compact media.
k 0 (4) Two main operators are needed to apply the energy
criterion for the accuracy of signal reconstruction by wavelet
can provide both a uniform (including a uniform best) coefficients: multivariate analysis [5] and calculation of the octave
and a mean-square approximation. It all depends on how you energy spectrum [9]. The advantage of the octave spectrum is that
calculate the coefficients. it, like the Fourier spectrum, is invariant with respect to time shifts
The Haar basics attract the attention of specialists for two of stationary signals. The property of a multiple-scale analysis is
reasons: also possessed by some wavelets considered on the whole axis t (-
1. Reducing the number of coefficients required to ∞, ∞), for example, Shannon wavelets[10].
approximate (to a given accuracy) based on the total number of We transform the continuous signal f (x) to a discrete
binary segments. form-we represent it as a row vector containing n real numbers fi, i
2. The absence of "long" operations in the expression = 0, 1,..., n-1. In algorithms for fast wavelet transforms, integer
(4.). Used only addition, subtraction and shift. iterations of a single scaling operator Dσ (σ> 1) describing the
With the use of Haar wavelets, all the higher listed stretching are actually used [8]. Usually a scale σ = 2 is used, in
properties are preserved. which the maternal wavelet satisfies the identity
n 1
REFERENCES
[1] Whittaker E .. On the Functions which are represented by Expansions of the
Interpolation Theory // Proc. Roy. Soc. Edinburgh. V.35. pp.181-194. 1915.
[2] Kotelnikov V. About the capacity of "ether" and wire in telecommunications //
Figure 3: The plot of the selected section of the magnetic field Uspekhi Fizicheskikh Nauk. T.176. 2006. № 7. p. 762-770 (Annex).
[3] Jerry A. Shannon's theorem of counts, its various generalizations and restrictions
of square form // TIER. 1977. T.65. № 11. c.53-89.
Figure 4: Graph of the amplitude spectrum of the magnetic [4] Ptacek M. Digital television. Theory and technique. Translation from Czech. M.:
Radio and communication. 1990. 528 p.
[5] Malla S. Wavelets in signal processing. Trans. with English. Moscow: Mir, 2005.
672 p.
[6] Marchuk GI, Agoshkov V.I. Introduction to projection-grid methods. M.:
Science. 1981. 416 p.
[7] Svinyin S.F. Basic splines in the theory of signal samples. SPb: Science. 2003.
118 pp.
[8] Blatter K. Wavelet analysis. Fundamentals of the theory. M.: The techno sphere.
2006. 272 p.
[9] Ahmed N, Rao K. Orthogonal transformations in the processing of digital signals.
M.: Communication. 1980. 248 p.
[10] Modern Sampling Theory: Mathematics and Applications. Benedetto J., Ferreiro
P. / Springer Science and Business Media LLC. 2012. 158p.
field
Conclusion