A.V. Bashkirov1, O.N. Chirkov2, A.S. Demikhova3

1-3 FSBEI of HE “Voronezh State Technical University” (Voronezh, Russia)

1 fabi7@mail.ru; 2 chir_oleg@mail.ru; 3 dem12.86@mail.ru

Problem statement. The paper considers deep learning (DNN) for processing OFDM wireless channels. In existing OFDM receivers, the channel status information (CSI) is first evaluated, and then the transmitted symbols are detected/restored using the evaluated CSI. A method based on deep learning is proposed, in which CSI is evaluated implicitly and transmitted characters are directly restored. To eliminate channel interference, the DNN model is first trained offline using the data obtained as a result of modeling based on channel statistics, and then used directly to restore the transmitted data.

Purpose. To consider the use of neural networks with deep learning DNN (Deep Neural Networks) for the complex task of evaluating the communication channel and detecting OFDM symbols.

Results. The proposed method based on deep learning makes it possible to eliminate channel distortions and detect transmitted characters with performance comparable to the estimation of the minimum standard deviation (MMSE). It is more effective than traditional assessment methods with an SNR exceeding 10 dB, when fewer training pilot signals (8 pilots) are used, there is no cyclic prefix and there is non-linear limiting noise.

Practical significance. Deep learning is a promising tool for channel estimation and signal detection in wireless communications with complex channel distortions and interference. The implementation of a deep learning neural network (DNN) of five levels is superior in performance to the least squares (LS) method and the minimum standard deviation (MMSE).

Bashkirov A.V., Chirkov O.N., Demikhova A.S. Application of deep neural networks for the task of channel estimation and signal detection in systems with OFDM. Radiotekhnika. 2025. V. 89. № 7. P. 10−14. DOI: https://doi.org/10.18127/j00338486-202507-02 (In Russian)

1. Chirkov O.N., Romashhenko M.A., Chepelev M.Ju. Sovremennye metody ocenki kanala radiosvjazi v uslovijah mnogoluchevosti. Vestnik Voronezhskogo gos. tehnich. un-ta. 2019. T. 15. № 3. S. 68-73. DOI: 10.25987/VSTU.2019.15.3.010 (in Russian).
2. Svidetel'stvo o gosudarstvennoj registracii programmy dlja JeVM № 2023686640 (RF). Programma modelirovanija ocenki besprovodnogo kanala svjazi s OFDM-moduljaciej na mashinnom obuchenii: № 2023685331. Chirkov O.N., Pirogov A.A.; zajavl. 22.11.2023; opubl. 07.12.2023 (in Russian).
3. Chirkov O.N., Pirogov A.A. Primenenie algoritmov mashinnogo obuchenija v zadache ocenki besprovodnogo kanala svjazi s OFDM. Vestnik Voronezhskogo gos. tehnich. un-ta. 2023. T. 19. № 6. S.164-169. DOI: 10.36622/VSTU.2023.19.6.025 (in Russian).
4. Wang X., Gao L., Mao S., Pandey S. CSI-based fingerprinting for indoor localization: A deep learning approach. IEEE Trans. Veh. Technol. Jan. 2017. V. 66. № 1. Р. 763–776.
5. Nachmani E., Beery Y., Burshtein D. Learning to decode linear codes using deep learning. 54’th Annual Allerton Conf. on Commun. Control and Computing. Mouticello. IL. Sept. 2016.
6. Romashhenko M.A., Chirkov O.N., Churakov P.P. Jeffektivnyj algoritm ocenki kanala radiosvjazi s prostranstvennoj moduljaciej. Radiotehnika. 2021. T. 85. № 6. S. 52-56. DOI: 10.18127/j00338486-202106-09 (in Russian).
7. Kyosti P. IST-4-027756 WINNER II D1.1.2 v.1.1. WINNER II Channel Models. 2007.
8. Romashhenko M.A., Chirkov O.N., Churakov P.P. Usovershenstvovannyj metod ocenki kanala s iteracionnym podavleniem pomeh dlja mnogopol'zovatel'skih sistem MIMO-OFDM. Radiotehnika. 2019. T. 83. № 6(8). S. 150-155. DOI: 10.18127/j00338486-201906(8)-11 (in Russian).
9. Chirkov O.N., Bashkirov A.V. Povyshenie jeffektivnosti ocenki kanala pri vysokoskorostnoj peredache dannyh v podvodnoj akusticheskoj svjazi s OFDM. Radiotehnika. 2024. T. 88. № 7. S. 45-49. DOI: 10.18127/j00338486-202407-09 (in Russian).