AI-Based Filtering of Signals Affected by Random Noise

Abstract

Digital signal processing (DSP) is at the core of modern technology and is widely used in diverse fields, including security, communications, medicine, space exploration, and other areas that rely on digital technologies or radio signals. Today, signal processing techniques can be combined with artificial intelligence (AI), enabling systems to unleash their full potential and effectively exploit them. This work aims to develop a machine learning algorithm, called Adaptive Pulses System (APS), that combines signal processing tools with AI learning capabilities to build, learn, and optimally utilize digital filters to solve one of the most significant problems in wireless communications. External influences such as noise and fading are major problems in wireless satellite communications, negatively impacting transmission quality. The carrier-to-noise ratio (C/N) generally represents the severity of these influences, with a C/N ratio below zero representing extremely poor signal conditions where reliable data extraction is difficult or impossible. The APS algorithm aims to build and learn high-performance DSP filters capable of raising the C/N ratio below zero and achieving a signal processing gain of more than 10 dB.