Automatic Digital Modulation Recognition Based on Novel Features and Support Vector Machine

In this paper a novel algorithm for automatic modulation recognition (AMR) based on pattern recognition approach is proposed. The main focus here remains on feature extraction block and the novel features are introduced in order to identify digital modulation schemes. The modulation types include: BASK, BFSK, BPSK, 4-ASK, 4-FSK, QPSK, and 16-QAM and the channel model is considered as an AWGN channel. The features are extracted from the received signal that is considered in the time, frequency and wavelet domains. Also, to overcome the multiclass problem, a hierarchical structure is investigated based on binary support vector machine (SVM). The simulations demonstrate superior capabilities of the proposed features in accurately separating digitally modulated signals in an extremely noisy environment with very low SNR values. Accordingly, the minimum SNR for the perfect identification is proven to be -5 dB, and a final accuracy percentage of 98.15 has been obtained in -10 dB.