Mitigação do crosstalk e das não linearidades em modulações por troca de cores utilizando redes neurais

Abstract

The significant increase in connected devices, overloading Wi-Fi infrastructure, known as the Wi-Fi Spectrum Crunch, has become an emerging challenge. In response to this growing demand, a solution to this issue is the development of visible light communication. This alternative utilizes the visible spectrum for efficient data transmission, especially through LED lamps. However, the technology that uses visible light faces some challenges, such as the nonlinearities of the transmitting LEDs, interference, noise, among others. To mitigate these obstacles, various machine learning methods can be employed, one of which is artificial neural networks. Therefore, this work models the visible light channel for color shift keying modulation schemes, using the channel state information matrix that includes crosstalk interference, as well as all components of additive noise. Additionally, the research implements Feedforward neural networks and adjust their parameters to achieve a more efficient model. This network can be effectively trained to mitigate inherent channel and transmission problems. Performance evaluation, in turn, considers crucial metrics such as the accuracy rate at various signal-to-noise power ratios and the average accuracy of intermediate signal-to-noise power levels. The results achieved through tests and adjustments demonstrated greater efficacy of Artificial Neural Networks in decoding Color Shift Keying transmissions compared to traditional methods, thus showing that the high accuracy of the models makes them a possible path for the future of visible light communication. Therefore, this study contributed to the advancement of visible light communication by integrating color shift keying modulation and artificial neural networks to improve signal reception and decoding. By offering an innovative and efficient solution to the current challenges in data transfer, the research sought to promote the effectiveness of optical communications and provide valuable insights for the future development of this promising area.