Recognition of fish acoustic patterns at monitoring of freshwater ecosystems

Abstract

The basic sources of contamination and obstruction of reservoirs are cleared not enough sewer water of industrial and communal enterprises, large stock-raising complexes, wastes of production; upcast of water and railway transport; wastes of roughing-out of flax, pesticides and other. Сontaminents, getting in natural reservoirs, result in the quality changes of water, that, mainly, appear in the change of physical properties of water, in the change of chemical composition of water, in a presence floating substances on the surface of water and laying of them on the bottom of reservoirs. The increases of population, expansion of old and origin of new cities considerably increased entering of domestic flows internal reservoirs. Synthetic cleansers that is widely used in the way of life contaminate reservoirs in a yet greater degree. In the total the capacity of waters goes down for oxigenating, activity of bacteria that mineralize organic substances is paralysed. The unfavorable ecological state of many freshwater ecosystems inflicts substantial harm to the fish resources of reservoirs and puts under a threat possibility not only to develop fish industry, conducting fish artificially, but also simply to catch her. All of it stimulate to do events in relation to the improvement of the ecological state of fresh reservoirs. Voice vibrations are the important constituent of the ecological monitoring of the biota state of fresh reservoirs. Information is about formation of sound in a reservoir part of that is activity of fishes turns out by means of acoustic sensors, that farther yields to computer treatment. The modern methods of recognition of fish acoustic patterns are based on the standards of signals, with properties of average estimations, or on comparisons of acoustic signals with a standard. It is shown that for creation of standards, as a rule, executed: previous signal processing, extraction of features of acoustic signal. Acoustic signals that act from movable objects – fishes can change depending on objective external terms and physical state of reservoirs. The hard algorithms of recognition of acoustic patterns are characterized high probability of error. In this connection repressing are adaptive algorithms of recognition of acoustic patterns. In the process of forming of standards clarification of software comes true according to the features of acoustic signal. Realization of process of creation of standards allows to determine the measure of functional readiness of parameters and knowledge base for the decision of recognition tasks of acoustic signals. In the process of recognition the probability terms of the correct comparing are set to the standard, on default of that an algorithm stops to be executed and requires additional studies. It requires creation of standards that reflect the characteristic features of fish signals. Presently for authentication mostly choose such pattern of acoustic signals, as period length of signal fundamental wave. It can be determined or by the search of maximal value in an autocorrelation function, or by the search of minimum value in the function of mean value of difference of signal amplitudes, or by the search of difference of two maximal values in the sequence of going into detail wavelet-coefficients. It is shown that for the tasks of recognition of fish acoustic patterns, most exact and requiring the least studies there is presentation of acoustic signal as a set of sign vectors of frames. In detail methodologies of the period selection of fundamental wave of acoustic signal were analysed: SIFT, EFT-А and EFT-WT. Methodology of EFT-WT is characterized absence of the thresholds set in good time; by the rapid search of period of fundamental wave; by absence of dependence on a noise-level, as a certain range of frequencies is investigated. At the same time calculable complication of wavelet transform is relatively high, in this connection it is necessary optimization of calculation algorithms.