Performance Improvement: Modulation Recognition Scheme Using CVCNN-LSTM Method

The process for automatic modulation recognition (AMR) is important to industrial cognitive radio Models. Although they have made significant strides in modulation detection, state-of-the-art real-valued CNNs compromise model interpretability and signal phase integrity by treating complicated signal components as distinct actual inputs. The CVCNN-LSTM is a novel AMR technique that is presented in this research. In order to capitalize on the advantages associated with both networks while minimizing their drawbacks, this study intends to use CNNs along with a LSTM. This method improves recognition performance by strengthening the learning for the initial in-phase/quadrature information (I/Q) signal characteristics. The suggested network consists of one LSTM layer and many local features learning unit (LFLUs). Convolutional and average pooling layers, which are the main components of an LFLU, are essential for allowing hierarchical correlation extraction and local correlation learning. Researchers recommend employing end-to-end complex-valued CNN & LSTM for modulation recognition rather than dual-channel real-valued networks as in-phase/quadrature information naturally including complex-valued system. Researchers analyse complex-valued networks via experiments and show that this methodology outperforms the state-of-the-art techniques in AMR area.