SCIENCE CHINA Information Sciences, Volume 63 , Issue 6 : 160303(2020) https://doi.org/10.1007/s11432-020-2851-0

AI based on frequency slicing deep neural network for underwater visible light communication

More info
  • ReceivedJan 10, 2020
  • AcceptedMar 24, 2020
  • PublishedMay 9, 2020



This work was partially supported by National Key Research and Development Program of China (Grant No. 2017YFB0403603) and Natural National Science Foundation of China (Grant No. 61925104).


[1] Zeng Z Q, Fu S, Zhang H H, et al. A survey of underwater optical wireless communications. IEEE Commun Surv Tut, 2017, 19: 204--238. Google Scholar

[2] Chi N, Haas H, Kavehrad M. Visible light communications: demand factors, benefits and opportunities [Guest Editorial]. IEEE Wireless Commun, 2015, 22: 5-7 CrossRef Google Scholar

[3] Zhou Y, Zhu X, Hu F. Common-anode LED on a Si substrate for beyond 15??Gbit/s underwater visible light communication. Photon Res, 2019, 7: 1019-1029 CrossRef Google Scholar

[4] Zhao Y, Zou P, Yu W. Two tributaries heterogeneous neural network based channel emulator for underwater visible light communication systems. Opt Express, 2019, 27: 22532-22541 CrossRef PubMed ADS Google Scholar

[5] 徐至展 . Nonlinear adaptive filters for high-speed LED based underwater visible light communication [Invited]Nonlinear adaptive filters for high-speed LED based underwater visible light communication [Invited]Nonlinear adaptive filters for high-speed LED based underwater visible light communication [Invited]. Chin Opt Lett, 2019, 17: 100011 CrossRef ADS Google Scholar

[6] Chi N, Zhao Y, Shi M. Gaussian kernel-aided deep neural network equalizer utilized in underwater PAM8 visible light communication system. Opt Express, 2018, 26: 26700-26712 CrossRef PubMed ADS Google Scholar

[7] Wu F M, Lin C T, Wei C C. Performance Comparison of OFDM Signal and CAP Signal Over High Capacity RGB-LED-Based WDM Visible Light Communication. IEEE Photonics J, 2013, 5: 7901507-7901507 CrossRef ADS Google Scholar

[8] Ziemer R E, Tranter W H. Principles of communications. John Wiley Sons, 2014. Google Scholar

[9] Zibar D, Piels M, Jones R. Machine Learning Techniques in Optical Communication. J Lightwave Technol, 2016, 34: 1442-1452 CrossRef ADS Google Scholar

[10] Khan F N, Lu C, Lau A P T. Machine learning methods for optical communication systems. In: Proceedings of Signal Processing in Photonic Communications, 2017. 3. Google Scholar

[11] Li G, Hu F, Zhao Y, et al. Enhanced performance of a phosphorescent white LED CAP 64QAM VLC system utilizing deep neural network (DNN) post equalization. In: Proceedings of IEEE/CIC International Conference on Communications in China (ICCC), Changchun, 2019. 173--176. Google Scholar

[12] Osahon I N, Rajbhandari S, Popoola W O. Performance Comparison of Equalization Techniques for SI-POF Multi-Gigabit Communication With PAM- M and Device Non-Linearities. J Lightwave Technol, 2018, 36: 2301-2308 CrossRef ADS Google Scholar

[13] Kaushal H, Kaddoum G. Underwater Optical Wireless Communication. IEEE Access, 2016, 4: 1518-1547 CrossRef Google Scholar

[14] Ali M A A, Mohammed M A. Effect of atmospheric attenuation on laser communications for visible and infrared wavelengths. Al-Nahrain J Sci, 2013, 16: 133--140. Google Scholar

[15] Johnson L, Green R, Leeson M. A survey of channel models for underwater optical wireless communication. In: Proceedings of 2013 2nd International Workshop on Optical Wireless Communications (IWOW), 2013. 1--5. Google Scholar

[16] 徐至展 . Recent achievements on underwater optical wireless communication [Invited]Recent achievements on underwater optical wireless communication [Invited]Recent achievements on underwater optical wireless communication [Invited]. Chin Opt Lett, 2019, 17: 100009 CrossRef ADS Google Scholar

[17] Huang X X, Wang Z X, Shi J Y, et al. 1.6 Gbit/s phosphorescent white LED based VLC transmission using a cascaded pre-equalization circuit and a differential outputs PIN receiver. Opt Express, 2015, 23: 22034--22042. Google Scholar

[18] Kim J, Konstantinou K. Digital predistortion of wideband signals based on power amplifier model with memory. Electron Lett, 2001, 37: 1417-1418 CrossRef Google Scholar

[19] Ju C, Liu N, Chen X. SSBI Mitigation in A-RF-Tone-Based VSSB-OFDM System With a Frequency-Domain Volterra Series Equalizer. J Lightwave Technol, 2015, 33: 4997-5006 CrossRef ADS Google Scholar

[20] Zhang J, Yu J, Li F. 11 5 9.3Gb/s WDM-CAP-PON based on optical single-side band multi-level multi-band carrier-less amplitude and phase modulation with direct detection.. Opt Express, 2013, 21: 18842-18848 CrossRef PubMed ADS Google Scholar

[21] Burse K, Yadav R N, Shrivastava S C. Channel Equalization Using Neural Networks: A Review. IEEE Trans Syst Man Cybern C, 2010, 40: 352-357 CrossRef Google Scholar

[22] Zhou Y, Zhang J, Wang C, et al. A novel memoryless power series based adaptive nonlinear pre-distortion scheme in high speed visible light communication. In: Proceedings of Optical Fiber Communication Conference and Exhibition (OFC), 2017. Google Scholar

[23] Haykin S O. Neural Networks and Learning Machines. Upper Saddle River: Pearson, 2009. 3. Google Scholar

  • Figure 1

    (Color online) (a) The frequency response of received CAP signal (Rx) and transmitted CAP signal (Tx) with the bandwidth of 450 MHz in our UVLC system. The constellation comparison in the case (b) with nonlinear effect and (c) without nonlinear effect in the UVLC system (w: with. w/o: without).

  • Figure 2

    (Color online) The schematic of FSDNN.

  • Figure 3

    (Color online) The frequency spectrum of recovered signal when (a) sps = 6, (b) sps = 7, (c) sps = 8.

  • Figure 4

    (Color online) Experimental setup. AWG: arbitrary wave generator. EA: electrical amplifier. Eq.: Equalizer. TIA: trans-impedance amplifier. OSC: oscilloscope. LPF: low pass filter. HPF: high pass filter. LFSDNN: the sub-FSDNN with the low-pass filter. HFSDNN: the sub-DNN with the high-pass filter.

  • Figure 5

    (Color online) BER results versus epoch when (a) DNN has one or two hidden layers, (b) FSDNN has one or two hidden layers.

  • Figure 6

    (Color online) BER performance versus different taps of (a) DNN and (b) FSDNN, including HFSDNN and LFSDNN. Taps is the number of input nodes in input layer.

  • Figure 7

    (Color online) BER performance versus different number of nodes in hidden layer of (a) DNN and (b) FSDNN.

  • Figure 8

    (Color online) BER performance versus different (a) taps and (b) step size ($u$) in the 2nd-stage LMS equalizer.

  • Figure 9

    (Color online) BER performance versus (a) different bias current and (b) Vpp for LMS, Volterra, DNN plus LMS and FSDNN plus LMS under the optimal structure. Insets: the constellation of the equalized signal using FSDNN at the certain Vpp of (i) 0.5 V, (ii) 0.7 V, (iii) 0.9 V and at the certain current of (i) 115 mA, (ii) 135 mA, (iii) 155 mA.

  • Figure 10

    (Color online) The structure of optimal DNN and FSDNN. I: input layer. H: hidden layer. O: output layer.