Skip to main content
placeholder image

Adaptive Extreme Learning Machine-Based Nonlinearity Mitigation for LED Communications

Journal Article


Abstract


  • This work concerns the receiver design for light emitting diode (LED) communications, where the LED nonlinearity can severely degrade the system performance. The LED nonlinearity makes high speed LED communications more challenging when it is time-varying (e.g., due to temperature drifting) and/or combined with time-varying channels (due to the relative movement between the transmitter and the receiver). In this work, we use adaptive neural network techniques to address this issue. We first propose a new adaptive extreme learning machine (AELM) with a variable forgetting factor for adaptive learning in dynamic scenarios. Then, an AELM based (turbo) receiver is designed to handle the time-varying LED nonlinearity and memory effects jointly. It is demonstrated that the proposed AELM based receiver can efficiently mitigate the dynamic nonlinearity and memory effects, and outperform the state-of-the-art adaptive techniques significantly.

Publication Date


  • 2021

Citation


  • Gao, D., Guo, Q., Jin, M., Yu, Y., & Xi, J. (2021). Adaptive Extreme Learning Machine-Based Nonlinearity Mitigation for LED Communications. IEEE Journal of Selected Topics in Quantum Electronics, 27(2). doi:10.1109/JSTQE.2020.3043779

Scopus Eid


  • 2-s2.0-85098797992

Volume


  • 27

Issue


  • 2

Abstract


  • This work concerns the receiver design for light emitting diode (LED) communications, where the LED nonlinearity can severely degrade the system performance. The LED nonlinearity makes high speed LED communications more challenging when it is time-varying (e.g., due to temperature drifting) and/or combined with time-varying channels (due to the relative movement between the transmitter and the receiver). In this work, we use adaptive neural network techniques to address this issue. We first propose a new adaptive extreme learning machine (AELM) with a variable forgetting factor for adaptive learning in dynamic scenarios. Then, an AELM based (turbo) receiver is designed to handle the time-varying LED nonlinearity and memory effects jointly. It is demonstrated that the proposed AELM based receiver can efficiently mitigate the dynamic nonlinearity and memory effects, and outperform the state-of-the-art adaptive techniques significantly.

Publication Date


  • 2021

Citation


  • Gao, D., Guo, Q., Jin, M., Yu, Y., & Xi, J. (2021). Adaptive Extreme Learning Machine-Based Nonlinearity Mitigation for LED Communications. IEEE Journal of Selected Topics in Quantum Electronics, 27(2). doi:10.1109/JSTQE.2020.3043779

Scopus Eid


  • 2-s2.0-85098797992

Volume


  • 27

Issue


  • 2