OFDM Channel Estimation Using Deep Learning Based on Convolutional Neural Networks

The problem of channel estimation in orthogonal frequency multiplexing (OFDM) systems using deep learning methods based on convolutional neural networks is considered. The paper compares the initial software implementation of classical channel estimation methods, and a modified implementation supplemented by neural network models such as FSRCNN (Fast Super‑Resolution Convolutional Neural Network – fast convolutional neural network for super-resolution). LS (least squares), practical LMMSE (with minimum standard deviation), ideal LMMSE (optimal LMMSE), as well as two convolutional models are considered: basic – CNN Base and advanced – CNN Deep. It is established that neural network models provide higher accuracy of channel estimation in almost the entire studied range of SNR values. The results obtained are of practical interest for designing digital paths of radio electronic equipment for broadband communication systems.

1. Чирков О.Н. Многополосный преобразователь частоты OFDM / О.Н. Чирков, Р.К. Астрединов // Проблемы обеспечения надежности и качества приборов, устройств и систем: Межвузовский сборник научных трудов. – Воронеж: Воронежский государственный технический университет, 2018. – С. 120–124.

2. Преображенский А.П. Особенности технологии OFDMA / А.П. Преображенский, О.Н. Чопоров // Вестник Воронежского института высоких технологий. – 2017. – Т. 11, № 4 (23). – С. 81–84.

3. Чирков О.Н. Оценка канала связи для OFDM систем с использованием методов глубокого обучения / О.Н. Чирков, Д.С. Мацокин, А.Г. Халдобин // Вестник Воронежского государственного технического университета. – 2025. – Т. 21, № 2. – С. 177–181.

4. Efficient Channel Estimation in OFDM Systems Using a Fast Super-Resolution CNN Model / S. Khichar, W. Santipach, L. Wuttisittikulkij [et al.] // Journal of Sensor and Actuator Networks. – 2024. – Vol. 13, No. 5. – URL: https://doi.org/10.3390/jsan13050055 (дата обращения: 16.04.2026).