Towards Enhancing Spectrum Sensing: Signal Classification Using Autoencoders

The demand for technologies relying on the radio spectrum, such as mobile communications and IoT, has been growing exponentially. As a consequence, providing access to spectrum is becoming increasingly more important. ever-growing wireless traffic increasing scarcity of available warrants efficient management spectrum. At same time, machine learning (ML) ubiquitous found applications in many fields its ability identify patterns assist with decision-making processes. Recently, algorithms have used address challenges domain, sensing, shown better performance than traditional sensing methods, energy detection. Spectrum method detecting identifying different signals being transmitted band crucial improving dynamic sharing, which potential enhance sharing coexistence frequency ultimately improve efficiency. To this end, research evaluates types autoencoders, deep, variational Long Short-Term Memory (LSTM) differentiate between LTE Wi-Fi transmissions. goal investigate autoencoders an I/Q dataset consisting combination (IEEE 802.11ax IEEE 802.11ac) classification task terms complexity, precision, recall best algorithm. Our models achieved up 99.9% precision 88.1% task. Additionally, shortest training time approximately 47 seconds, are suitable online deployment RF environment.