Dual-Stream CNN-BiLSTM Model with Attention Layer for Automatic Modulation Classification

The Automatic Modulation Classification (AMC) approach is utilized to determine the nature of modulation. It is a significant job for intelligent receivers, essential components of future wireless communication technologies, including adaptive modulation and cognitive radio. Deep learning is often utilized for automatic modulation classification because of its superior performance and learning capability. A dual-stream CNN and BiLSTM architecture are used in this paper to classify seven different forms of digital modulation (8PSK, CPFSK, GFSK, QAM64, QPSK, BPSK, QAM16). Additionally, an attention layer is added to the model to enhance performance. Moreover, using dual streams rather than a single one improves the quality of the output. The proposed model is compared with earlier works. In comparison to the current research, it is observed that the proposed model provides superior classification accuracy with less complexity. The time needed for classification is also decreased due to the reduction in complexity.