AUTOMATIC MODULATION CLASSIFICATION USING DEEP LEARNING POLAR FEATURE

The automatic modulation classification of signals is great importance in modern communications, especially on cognitive radio. Several methods have been used this field, the most important which automatically using Deep Learning, where depend convolution neural network, one Learning networks, achieved high accuracy classifying modulation, so proposed network depends type deep learning CNN consisting four blocks, each block contains a set symmetric and asymmetric filters. also Max Pool. In paper, features extracted phase-squaring polar combined for input, helps extending that is, an increase inside network. It contributes to improving higher-order through Polar plane. dataset RadioML 2018.01A was adopted, recent research, 11 types normal-class: (FM, GMSK, QPSK, BPSK, 0QPSK, AM-SSB-SC, 4ASK, AM-DSB-SC, 16QAM, 8PSK,00K) were taken. A simulation can be found Matlab 2021. 100% when signal-to-noise ratio greater or equal 2 dB modulation. results paper compared with networks Baseline Visual Geometry Group Residual Neural comparison showed superiority over these as at SNR while BL 72% dB, RN, VGG almost reach 93% dB.