Спектрален анализ и идентификация на шумове чрез метод на опорните вектори и невронно-размити класификатори
Noise Spectral Analysis and Identification by Support Vector Machine and Neuro-Fuzzy Classifiers

Summary/Abstract: The report implements hybrid research on spectral feature extraction and synthesis of Gaussian and Periodic noise recognition models to analog and digital signals in communication link channels. Frequency analysis is based on Fast Fourier Transform when setting Hamming and Hanning window functions. Procedures for training and evaluating the performance of models against the specified spectral categories, created on the basis of Support Vector Machine (SVM) and Adaptive Neuro-Fuzzy Interface Systems (ANFIS), are carried out. In SVM methods, classification accuracy is analyzed for selected Kernel functions Linear, Polynomial, Radial Basis and Sigmoid. The suitability of ANFIS classification models is investigated in two directions - a combined approach with the Method of Least Squares and the Method of Backpropagation of the Error.

• Details
• Contents