Developing and Building an Intelligent Inter-Symbol Interference Reduction System Using Deep Learning Techniques

Multipath fading and inter-symbol interference are two common issues that can arise in digital communication systems. These issues can significantly affect the quality of signal transmission and can cause significant distortion of the original signal. In this paper, we propose an improved model to mitigate (cancel) the inter-symbols interference which is one of multipath fading problem effects by building a deep learning model using recurrent neural networks to predict the coefficient values that help in signal recovery. We compared the performance of four equalizers based on neural network, Least Mean Square (LMS) Equalizer, Recursive least squares (RLS), Decision Feedback Equalizer (DFE) based on Recursive least squares algorithm (RLS), and Decision Feedback Equalizer (DFE) based on Least Mean Square algorithm (LMS). In simulation, the new DFE RLS equalizer outperforms the other previously mentioned equalizers. The order from best to worst performance is: DFE RLS equalizer, DFE LMS equalizer, RLS then LMS. We used QAM (Quadrature Amplitude Modulation) for modulation.