Modulation recognition of terahertz signals by deep learning

The availability of wireless communication at the terahertz frequency band has emerged as a key research field for the next generation of networks, as 5G spectrum resources become increasingly limited due to the rise in the number of wireless connections. However, the burden on base-band signal processing is increased because of the erratic fluctuations in signal waveforms caused by elements like severe weather and urban multipath scattering. Therefore, a method is needed to reduce this burden and improve processing efficiency. This work investigates the recognition of signal modulation schemes by employing deep learning under various SNR situations. It explores two networks of CNN and LSTM, and establishes both indoor and outdoor environments to accomplish effective signal processing in severe weather.