Research Of Machine Learning Algorithm For Broadcasting Spectrum Signal Processing

FM broadcasting is mainly used to transmit sound and other signals in the form of wireless transmission.It has many characteristics such as wide coverage,easy installation,high sound quality,low cost,etc.Compared with traditional cable broadcasting,it has irreplaceable advantages.This thesis is based on a project for a radio department,and mainly focuses on the broadcasting frequency band(87 MHz to 108 MHz),uses radio monitoring equipment to scan the spectrum signal of the broadcasting frequency band,and performs the data preprocessing,signal feature extraction and classification processing of the radio frequency spectrum signal on the extracted spectrum signal,so as to extract abnormal signals such as pseudo base stations,black radio,cheats in exams,etc.The main research work is as follows:Firstly,relevant pre-processing is performed on the spectrum information of the broadcast frequency band.Through the improved K-Means algorithm,the original sample data including the glitch signal is eliminated,and the original signal is processed through the wavelet analysis of the spectrum signal.The original signal was decomposed by wavelet and wavelet reconstruction,so as to achieve the purpose of denoising the original signal.Secondly,based on the analysis of signal characteristics and the comparison of a large number of spectrum signals,a method for extracting individual features of spectrum signals is summarized,and the overall signal characteristics of 11 types of broadcast frequency bands are extracted.Finally,the grey relational degree cluster Analysis is used to extract the features of the spectrum signal,which provides a certain basis for the subsequent classification algorithm.Further,according to the characteristics of broadcast frequency spectrum information,combined with the theory and experiment of support vector machine and extreme learning machine,a series of comparisons are made between SVM,ELM,and combined classifiers to find out that the hybrid combined classification algorithm applies to broadcast frequency spectrum information.The classification effect is better.Finally,a classification algorithm that combines the gradient boosting decision tree(GBDT)and logistic regression(LR)features is applied to the spectrum processing of broadcasting frequency band.Compared with other algorithms,it has very great advantages.