Chinese Optics, Volume. 16, Issue 2, 415(2023)
Spatial pulse position modulation multi-classification detector based on deep learning
Fig. 2. Bit error performance of a (2,3,4)-SPPM system with different detection methods
Fig. 3. Bit error performances of two SPPM systems with different detection methods
Fig. 4. Effect of turbulence on the bit error performance of the system
Fig. 5. Influence of PPM order on the system’s bit error performance
Table 1. Turbulence model parameters
View tableView in ArticleTable 1. Turbulence model parameters
湍流模型 G-G信道[ 11 ]EW信道（D=25 mm）[ 20 ]参数 ${\alpha _{\rm{G}}}$ $ {\beta _{\rm{G}}} $ $ \sigma _p^2 $ $ {\alpha _{\rm{E}}} $ $ {\beta _{\rm{E}}} $ $ \eta $ Rytov方差 弱湍流 11.6 10.9 0.2 3.67 1.97 0.73 0.317 中等湍流 4.0 1.9 1.6 5.37 0.81 0.33 2.202 强湍流 4.2 1.4 3.5 5.50 0.74 0.29 15.851
Table 2. Hyperparameters of the multi-classification detector
View tableView in ArticleTable 2. Hyperparameters of the multi-classification detector
超参数 值 各隐藏层神经元数目 F1=64,F2=98,F3=48 Batch 1.25×104 Batch_size 24 轮次Epoch 50 激活函数 Relu+Sigmoid 损失函数 Cross Entropy Loss 优化器 SGD 学习率 0.001
Table 3. Computational complexity of each algorithm
View tableView in ArticleTable 3. Computational complexity of each algorithm
检测算法 计算复杂度/Flops ML 检测 $ {N_t}L\left( {2{N_t}{N_r}L + 2{N_r}L - 1} \right) $ KMC分步分类检测[23] $ {N_t}\left( {2{N_t}{N_r}L + {\text{2}}{N_r}L - 1} \right) + L\left( {3{N_r}L - 1} \right) $ 线性均衡DNN检测器[17] $ 2\left( {{{\left( {{N_r}L} \right)}^2} + {N_r}L{F_1} + {F_1}{F_2} + {F_2}{F_3} + {F_3}{{\log }_2}\left( {{N_t}L} \right)} \right) + {\log _2}\left( {{N_t}L} \right) $ DNN多分类检测器 $ 2\left( {{N_r}L{F_1} + {F_1}{F_2} + {F_2}{F_3} + {F_3}L + N_t^2{N_r}L + {N_t}{N_r}L} \right) + 3L - {N_t} - 1 $
Get Citation
Copy Citation Text
Hui-qin WANG, Wen-bin HOU, Rui HUANG, Dan CHEN. Spatial pulse position modulation multi-classification detector based on deep learning[J]. Chinese Optics, 2023, 16(2): 415
Category: Original Article
Received: May. 27, 2022
Accepted: --
Published Online: Apr. 4, 2023
The Author Email:
