Deep Learning Models For Symbol Detection in UFMC Systems

The limited availability of the frequency band in wireless communication systems is one of the major obstacles to achieving high-speed data transmission. To overcome this obstacle, multicarrier systems, which utilize the available frequency bandwidth most efficiently to ensure spectral efficiency and consequently high data rate transmission, are used. In the Universal Filtered Multi-Carrier (UFMC) technique, which is one of the multi-carrier systems, in addition to high-speed data transmission, the bandwidth is divided into many sub-bands and only the lower sidebands are filtered, and as a result, the inter-channel interference problem is minimized. However, in UFMC systems, the error-free reception of symbols at the receiver is directly dependent on the performance of the symbol detection algorithm. In this study, symbol detection was performed in UFMC systems by taking advantage of the learning ability of deep learning methods, providing flexible solutions in solving nonlinear problems, reducing the hardware load by using fewer parameters and the ability to perform parallel processing, and thus the symbol detection performance of the system under bad channel conditions was increased.