Intelligent Recognition for Fast Access to Machine to Machine

A key technology of Machine to Machine (M2M) communication system is link adaptation. The identification of modulation mode is of great significance to link adaptation. The traditional automatic modulation classification (AMC) method uses support vector machine (SVM) to classify signals, which has the following problems. The SVM classifier is a supervised method and the computation is huge. Aiming at the problems of traditional AMC methods, this paper creatively proposes a modulation classification method based on Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. The method uses machine learning, and the computational load is far less than the deep learning used by the traditional AMC algorithm. In this paper, 5 feature selection methods are used to extract 12 parameters and classify signals. 5 feature selection methods are compared and their contributions to classification are analyzed. The experimental results show the excellent accuracy of modulation classification when the SNR is 5dB.