DEEP LEARNING-BASED DECODING FOR PHASE SHIFT KEYING

In this paper, the application of Deep Learning (DL) in the field of telecommunications is discussed, focusing specifically on symbol detection at the receiver. The performance of Deep Learning-based detection is examined for phase shift keying modulation over Additive White Gaussian Noise (AWGN) and Rayleigh channels. First, a model is proposed which shows that the theoretical bit error rate and throughput can be achieved using DL techniques. Then, the effects of different DL model parameters on the model performance are investigated. The DL model for symbol detection with tuned and minimized parameter set is examined from various aspects and it is shown that this improved version can achieve the desired results with much less complexity of the realization.