Design of a Deep Learning based Intelligent Receiver for a Wireless Communication System

In communication systems, deep learning techniques can provide better predictions than model-based methods when the hidden features of the problem are prone to deviating substantially from the formulated assumptions. Severe signal impairments due to multipath fading and higher channel noise levels degrade the performance of conventional receivers. To overcome this, a novel intelligent receiver based on a deep learning network is presented, achieving better performance in terms of reduced bit error rate than a standalone conventional receiver. The experimental result shows that the relative decrement in the symbol error ratio due to the proposed method is about 9 percent compared to the traditional receiver when the Rician channel fading is relatively high.