Abstract
Machine learning in the physical layer of communication systems holds the potential to improve performance and simplify design methodology. Many algorithms have been proposed; however, the model complexity is often unfeasible for real-time deployment. The real-time processing capability of these systems has not been proven yet. In this work, we propose a novel, less complex, fully connected neural network to perform channel estimation and signal detection in an orthogonal frequency division multiplexing system. The memory requirement, which is often the bottleneck for fully connected neural networks, is reduced by ≈ 27 times by applying known compression techniques in a three-step training process. Extensive experiments were performed for pruning and quantizing the weights of the neural network detector. Additionally, Huffman encoding was used on the weights to further reduce memory requirements. Based on this approach, we propose the first field-programmable gate array based, real-time capable neural network accelerator, specifically designed to accelerate the orthogonal frequency division multiplexing detector workload. The accelerator is synthesized for a Xilinx RFSoC field-programmable gate array, uses small-batch processing to increase throughput, efficiently supports branching neural networks, and implements superscalar Huffman decoders.
- [1] GitHub. n.d. google-research/google-research. Retrieved November 8, 2021 from https://github.com/google-research/google-research.Google Scholar
- [2] S. Yue, Y. Ma, N. Yi, and R. Tafazolli. 2018. Unsupervised deep learning for MU-SIMO joint transmitter and noncoherent receiver design. IEEE Wireless Communications Letters 8, 1 (2018), 177–180. https://doi.org/10.1109/LWC.2018.2865563Google Scholar
- [3] IEEE. 2019. 754-2019—IEEE Standard for Floating-Point Arithmetic. IEEE, Los Alamitos, CA. https://doi.org/10.1109/IEEESTD.2019.8766229Google Scholar
- [4] . 2019. 3GPP TR 21.915.
Technical Report 3GPP TR 21.915 V15.0.0 (2019-09). Retrieved November 8, 2021 from https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=3389.Google Scholar - [5] . 2020. 3GPP TR 36.912.
Technical Report 3GPP TR 36.912 V16.0.0 (2020-07-14). Retrieved November 8, 2021 from https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=2584.Google Scholar - [6] . 2021. 3GPP TR 36.211.
Technical Report 3GPP TR 36.211 V16.6.0 (2021-06-30). Retrieved November 8, 2021 from https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=2425.Google Scholar - [7] . 2014. Gzip on a chip: High performance lossless data compression on FPGAs using OpenCL. In Proceedings of the International Workshop on OpenCL 2013 & 2014 (IWOCL’14). ACM, New York, NY, 1–9. https://doi.org/10.1145/2664666.2664670 Google Scholar
Digital Library
- [8] . 2018. Neural Networks and Deep Learning: A Textbook. Springer International, Cham, Switzerland. https://doi.org/10.1007/978-3-319-94463-0 Google Scholar
Digital Library
- [9] . 2000. Parallel Huffman decoder with an optimized look up table option on FPGA. In Proceedings of the 2000 TENCON Proceedings on Intelligent Systems and Technologies for the New Millennium (Cat. No. 00CH37119), Vol. 1. https://doi.org/10.1109/TENCON.2000.893543Google Scholar
Cross Ref
- [10] . 2017. An OpenCL deep learning accelerator on Arria 10. In Proceedings of the 2017 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays (FPGA’17). ACM, New York, NY, 55–64. https://doi.org/10.1145/3020078.3021738Google Scholar
Digital Library
- [11] . 2007. OFDM Baseband Receiver Design for Wireless Communications. John Wiley & Sons.
Google Scholar Digital Library
- [12] B. C. Csáji and others. 2001. Approximation with artificial neural networks. Faculty of Sciences 24, 48 (2001), 7.Google Scholar
- [13] . 2018. Deep learning based communication over the air. IEEE Journal of Selected Topics in Signal Processing 12, 1 (
Feb. 2018), 132–143. https://doi.org/10.1109/JSTSP.2017.2784180Google ScholarCross Ref
- [14] . 2009. FPGA implementation of an improved channel estimation algorithm for mobile WiMAX. In Proceedings of the 2009 International Conference on Microelectronics (ICM’09). 280–283. https://doi.org/10.1109/ICM.2009.5418629
Google Scholar Cross Ref
- [15] . 2018. OFDM-autoencoder for end-to-end learning of communications systems. In Proceedings of the 2018 IEEE 19th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC’18). 1–5. https://doi.org/10.1109/SPAWC.2018.8445920Google Scholar
Cross Ref
- [16] . 2018. ComNet: Combination of deep learning and expert knowledge in OFDM receivers. IEEE Communications Letters 22, 12 (
Dec. 2018), 2627–2630. https://doi.org/10.1109/LCOMM.2018.2877965Google ScholarCross Ref
- [17] . 2016. Deep Learning. MIT Press, Cambridge, MA. Google Scholar
Digital Library
- [18] . 2018. A survey on methods and theories of quantized neural networks. arXiv:1808.04752 [cs, stat] (
Dec. 2018). http://arxiv.org/abs/1808.04752.Google Scholar - [19] . 2016. EIE: Efficient inference engine on compressed deep neural network. arXiv:1602.01528 [cs] (
Feb. 2016). http://arxiv.org/abs/1602.01528.Google Scholar - [20] . 2015. Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding. arXiv:1510.00149 [cs] (
Oct. 2015). http://arxiv.org/abs/1510.00149.Google Scholar - [21] . 2020. haoyye/OFDM_DNN. Retrieved November 8, 2021 from https://github.com/haoyye/OFDM_DNN.
Google Scholar - [22] . 2019. Model-driven deep learning for physical layer communications. IEEE Wireless Communications 26, 5 (
Oct. 2019), 77–83. https://doi.org/10.1109/MWC.2019.1800447Google Scholar Cross Ref
- [23] 2018. Zynq UltraScale+ RFSoC Data Sheet: Overview. Retrieved November 8, 2021 from https://www.xilinx.com/support/documentation/data_sheets/ds889-zynq-usp-rfsoc-overview.pdf.Google Scholar
- [24] 2019. DPU for Convolutional Neural Network v3.0, DPU IP Product Guide. Retrieved November 8, 2021 from https://www.xilinx.com/support/documentation/ip_documentation/dpu/v3_0/pg338-dpu.pdf.Google Scholar
- [25] 2020. UltraScale Architecture-Based FPGAs Memory IP v1.4 LogiCORE IP Product Guide. Retrieved November 8, 2021 from https://www.xilinx.com/support/documentation/ip_documentation/ultrascale_memory_ip/v1_4/pg150-ultrascale_memory-ip.pdf.Google Scholar
- [26] . 2017. Quantization and training of neural networks for efficient integer-arithmetic-only inference. arXiv:1712.05877 [cs, stat] (
Dec. 2017). http://arxiv.org/abs/1712.05877.Google Scholar - [27] . 2021. Equalize OFDM Data Using Channel Estimates. Retrieved November 8, 2021 from https://uk.mathworks.com/help/wireless-hdl/ref/ofdmequalizer.html.Google Scholar
- [28] . 2021. Estimate Channel Using Input Data and Reference Subcarriers. Retrieved November 8, 2021 from https://uk.mathworks.com/help/wireless-hdl/ref/ofdmchannelestimator.html.Google Scholar
- [29] . 2012. Real-Time Systems Design and Analysis: Tools for the Practitioner (4th ed.). Wiley. http://www.ezproxy.is.ed.ac.uk/login?url=http://onlinelibrary.wiley.com/book/10.1002/9781118136607. Google Scholar
Digital Library
- [30] . 2020. High-throughput FPGA-based hardware accelerators for deflate compression and decompression using high-level synthesis. IEEE Access 8 (2020), 62207–62217. https://doi.org/10.1109/ACCESS.2020.2984191
Google Scholar Cross Ref
- [31] . 2019. The roadmap to 6G: AI empowered wireless networks. IEEE Communications Magazine 57, 8 (
Aug. 2019), 84–90. https://doi.org/10.1109/MCOM.2019.1900271Google Scholar Digital Library
- [32] . 2015. Flexible configured OFDM for 5G air interface. IEEE Access 3 (2015), 1861–1870. https://doi.org/10.1109/ACCESS.2015.2480749
Google Scholar Cross Ref
- [33] n.d. ADM-XRC-9R1: FPGA COTS board: Xilinx Zynq —8x 14-bit 4/5Gsps ADCs, 8x 14-bit 6.5/10Gsps DACs. Retrieved November 8, 2021 from https://www.alpha-data.com/esp/products.php?product=adm-xrc-9r1.Google Scholar
- [34] . 2018. ALAMO: FPGA acceleration of deep learning algorithms with a modularized RTL compiler. Integration, the VLSI Journal 62 (
June 2018), 14–23. https://doi.org/10.1016/j.vlsi.2017.12.009Google ScholarDigital Library
- [35] . 2018. Deep learning for intelligent wireless networks: A comprehensive survey. IEEE Communications Surveys Tutorials 20, 4 (2018), 2595–2621. https://doi.org/10.1109/COMST.2018.2846401Google Scholar
Digital Library
- [36] . 2012. Wireless Communications. John Wiley & Sons.
Google Scholar Digital Library
- [37] . 2017. Deep learning based MIMO communications. arXiv:1707.07980 [cs, math] (
July 2017). http://arxiv.org/abs/1707.07980.Google Scholar - [38] . 2017. An introduction to deep learning for the physical layer. IEEE Transactions on Cognitive Communications and Networking 3, 4 (
Dec. 2017), 563–575. https://doi.org/10.1109/TCCN.2017.2758370Google ScholarCross Ref
- [39] . 2018. Modified Huffman based compression methodology for deep neural network implementation on resource constrained mobile platforms. In Proceedings of the 2018 IEEE International Symposium on Circuits and Systems (ISCAS’18). 1–5. https://doi.org/10.1109/ISCAS.2018.8351234
Google Scholar Cross Ref
- [40] . 2016. Efficient deep neural network acceleration through FPGA-based batch processing. In Proceedings of the 2016 International Conference on ReConFigurable Computing and FPGAs (ReConFig’16). 1–8. https://doi.org/10.1109/ReConFig.2016.7857167Google Scholar
Cross Ref
- [41] . 2018. Throughput optimizations for FPGA-based deep neural network inference. Microprocessors and Microsystems 60 (
July 2018), 151–161. https://doi.org/10.1016/j.micpro.2018.04.004arXiv: 1810.00722. Google ScholarCross Ref
- [42] . 2016. Going deeper with embedded FPGA platform for convolutional neural network. In Proceedings of the 2016 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays (FPGA’16). ACM, New York, NY, 26–35. https://doi.org/10.1145/2847263.2847265 Google Scholar
Digital Library
- [43] . 2011. LTE—The UMTS Long Term Evolution: From Theory to Practice. John Wiley & Sons.
Google Scholar Digital Library
- [44] . 2019. FPGA-based accelerators of deep learning networks for learning and classification: A review. IEEE Access 7 (2019), 7823–7859. https://doi.org/10.1109/ACCESS.2018.2890150Google Scholar
Cross Ref
- [45] . 2019. ViterbiNet: Symbol detection using a deep learning based Viterbi algorithm. In Proceedings of 2019 IEEE 20th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC’19). 1–5. https://doi.org/10.1109/SPAWC.2019.8815457
Google Scholar Cross Ref
- [46] . 1976. On the construction of Huffman trees. In Third International Colloquium on Automata, Languages and Programming, University of Edinburgh, UK, July 20–23, 1976. Edinburgh University Press, Edinburgh, Germany, 382–410.Google Scholar
- [47] . 2017. Deep learning for wireless physical layer: Opportunities and challenges. China Communications 14, 11 (
Nov. 2017), 92–111. https://doi.org/10.1109/CC.2017.8233654Google ScholarCross Ref
- [48] . 2019. Artificial intelligence defined 5G radio access networks. IEEE Communications Magazine 57, 3 (
March 2019), 14–20. https://doi.org/10.1109/MCOM.2019.1800629 Google ScholarDigital Library
- [49] . 2018. Power of deep learning for channel estimation and signal detection in OFDM systems. IEEE Wireless Communications Letters 7, 1 (
Feb. 2018), 114–117. https://doi.org/10.1109/LWC.2017.2757490Google ScholarCross Ref
- [50] . 2018. A framework for generating high throughput CNN implementations on FPGAs. In Proceedings of the 2018 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays (FPGA’18). ACM, New York, NY, 117–126. https://doi.org/10.1145/3174243.3174265 Google Scholar
Digital Library
- [51] . 2015. Optimizing FPGA-based accelerator design for deep convolutional neural networks. In Proceedings of the 2015 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays (FPGA’15). 161–170. https://doi.org/10.1145/2684746.2689060 Google Scholar
Digital Library
- [52] . 2019. Deep learning in mobile and wireless networking: A survey. arXiv:1803.04311 [cs] (
Jan. 2019). http://arxiv.org/abs/1803.04311.Google Scholar - [53] . 2019. 6G Visions: Mobile ultra-broadband, super Internet-of-Things, and artificial intelligence. China Communications 16, 8 (
Aug. 2019), 1–14. https://doi.org/10.23919/JCC.2019.08.001Google Scholar Cross Ref
Index Terms
A Real-Time Deep Learning OFDM Receiver
Recommendations
FPGA Implementation of Subcarrier Index Modulation OFDM Transceiver
IPDPSW '13: Proceedings of the 2013 IEEE 27th International Symposium on Parallel and Distributed Processing Workshops and PhD ForumOFDM (Orthogonal frequency division multiplexing) is an efficient modulation scheme for wide-band digital communications and applications ranging from modems to next-generation high-speed wireless data communications[9]. Several enhancements of the OFDM ...
Performance Comparision of Autocorrelation and CPRDIC Algorithm Implemented on FPGA for OFDM Based WLAN
ICCSN '09: Proceedings of the 2009 International Conference on Communication Software and NetworksThis paper deals with design and implementation of Autocorrelator and CORDIC algorithm for OFDM based WLAN on FPGA. The autocorrelator is used for frame detection and carrier frequency offset estimation. The CORDIC is used to estimate the frequency ...
A Power-Line Communication Modem Based on OFDM
CONIELECOMP '09: Proceedings of the 2009 International Conference on Electrical, Communications, and ComputersIn this paper, we present the design and implementation of a PLC (Power-Line Communication) Modem based on Orthogonal Frequency Division Multiplexing (OFDM). The PLC device implements OFDM in both transmitter and receiver using VHDL programming. The ...






Comments