The analysis and use of mathematical methods for the detection of sound signals

Voice recognition is a widely studied and explored in science, and the recognition of audio signals, including cough patients in long audio recording is practically not studied. To recognize the audio signals and their counting were analyzed and used mathematical methods such as correlation analysis, fast Fourier transform, fuzzy logic. Looking at each method in turn, were selected for the parameters gives the best recognition results when using either method. Correlation analysis allows to draw conclusions about the number of cough moments for the selected standard. Fast Fourier transform allowed to allocate these frequency ranges, which are the only cough and do not get the noise was experimentally identified 6 of these ranges. The use of fuzzy logic has improved the analysis of sound recordings and allowed to make the choice of the cough moments with more certainty, but certainty of results to achieve and failed. Used fuzzy logic, which allows to expand the boundaries of recognition, allowing graded result in coughing may cough and noise. When analyzing the data obtained for each of the mathematical method, came to the conclusion that individually cannot be used. To achieve this goal it is necessary to compile a set of mathematical rules of inference that allow to recognize the sound signals with higher accuracy. Having considered the spectrogram of a sound clip, select the zone in which it is necessary to conduct an analysis to identify the similarity of sound recordings. The zones of cough have distinct values for each of the coordinates of the spectrogram. This allowed later when synthesizing inference rules use the exact values for the parameters of these rules. Under the resulting set of rules modified the developed software. Tested and received results.