Automatic Classification of Modulation Schemes under Blind Scenario

Automatic modulation classification (AMC) is getting progressively significant in spectrum monitoring and cognitive radio. At the receiver side, both the identification and demodulation of signal rely upon the right modulation classification. The created acoustic channels are submerged, usually seen as one of the most irksome correspondence sources being utilized today. The spread of the acoustic network is best supported at low frequencies, and the information transmission available for correspondence is very con-stressed. Due to the multifaceted nature and feebleness of acoustic correspondence structures created submerged, it is difficult to perceive modification during real correspondence. In this paper, we have proposed a novel technique to classify four modulation schemes including BPSK, CPFSK, DSB-AM, and GFSK. The CPFSK and GFSK are classified for the first time with analog modulation. Firstly, spectrograms are formed from the signals, and features extracted from signals and RGB channels of spectrograms are then fused serially. Secondly, Analysis of variance was incorporated to diminish unnecessary features to enhance the system's computational efficiency. The system outcome successfully achieved an accuracy of 99.6% on the linear support vector machine (L-SVM).