Return
Deep learning based repetition coding for MIMO VLC systems
DOI:10.1007/s11082-025-08317-7.png)
Abstract
En 中文
In this paper, an enhanced model of Optical Wireless Communications (OWC) is investigated utilizing Multiple Input Multiple Output (MIMO) strategies. The non-Lambertian light pillars, explicitly from economically accessible LEDs are applied to arrange MIMO-OWC joints in commonplace indoor conditions. Both homogeneous and heterogeneous non-Lambertian MIMO arrangements are examined. To assess the spatial convergence execution, a low-intricacy Redundancy Coding (RC) MIMO calculation is utilized. Moreover, combining the Single Shot Multi-Box Detector (SSD) model with the AdaBoost machine Learning (ML) model is implemented to enhance the performance and reliability of these techniques. In that context, a Multi-Path Feature Fusion Single Shot Multi-Box Detector (MF-SSD) is proposed. A new detection network is used with an effective feature fusion module based on SSD and DenseNet121 as a backbone. Two recently created modules with dilated convolution make up the suggested feature fusion module, which fuses features from shallow layers, which primarily contain boundary information, to higher-level features which primarily contain semantic reach information without lowering the feature map initial resolution. Furthermore, the model is fine-tuned to be suitable for our prediction task. Furthermore, K-fold cross validation is utilized to overcome the overfitting and evaluate the non- trainable parameters in the proposed framework. The MF-SSD model outperforms the SSD-AdaBoost model in all different scenarios typical to the kind of the transmitters.
Keywords:
Deep learning (DL)
Machine learning (ML)
Multiple input multiple output (MIMO)
Multi-path feature fusion single shot multi-box detector (MF-SSD)
Redundancy coding (RC)
Journal
IF:
4
Papers:
9.9K
Citations:
1.8W

