CNN-based AMR of simultaneously transmitted SC-QAM and OFDM-QAM signals in a 220 GHz terahertz system

The terahertz (THz) frequency band is characterized by continuous and exceptionally abundant available spectrum resources. Consequently, it is widely recognized as a pivotal frequency band for surmounting the rate bottlenecks of existing wireless communications and supporting the requirements of future ultra-high-speed communications. By incorporating band-pass delta-sigma modulation (BP-DSM) technology, THz communication systems enable the efficient transmission of high-order quadrature amplitude modulation (QAM) signals, thereby markedly enhancing overall system performance. In this context, THz communication systems must not only achieve a reasonable trade-off between energy and spectral efficiency but also support adaptive demodulation of QAM signals with diverse modulation orders at the receiver to satisfy the demand for flexibility in complex communication scenarios. This paper introduces a convolutional neural network (CNN) into a 220-GHz THz transmission system, enabling automatic modulation recognition (AMR) for simultaneously transmitted single-carrier QAM (SC-QAM) and orthogonal frequency-division multiplexing QAM (OFDM-QAM) signals ranging from 32QAM to 512QAM. The architecture of the recognition model comprises five convolutional layers and four max-pooling layers. By employing a design of progressively decreasing convolutional kernels, the model extracts deep abstract features from both SC-QAM and OFDM-QAM signals. Experimental results verify the recognition performance under nine different modulation combinations. The results show that the model achieves stable, high-performance modulation classification on OFDM-QAM signals. For SC-QAM signals, the model accurately identifies modulation levels from 32QAM to 256QAM, but recognition accuracy decreases for 512QAM. Meanwhile, the model demonstrates consistent robustness across varying wireless transmission distances. This study provides a vital theoretical foundation and experimental support for the development of intelligent, highly efficient wireless transmission systems operating in the THz frequency band.