Fig. 1. Single-scattering link model of UV non-line-of-sight (NLOS) communication

Fig. 2. LSTM-DNN based channel equalization scheme

Fig. 3. LSTM basic unit model

Fig. 4. DNN model. (a) DNN general model structure; (b) DNN single neuron model

Fig. 5. Normalized channel pulse response intensity curves for different parameters. (a) Transmitter-receiver distance; (b) beam divergence angle and receiving field of view angle; (c) transmitting and receiving elevation angle

Fig. 6. Effect of different geometric angles on transmission attenuation. (a) Transmitting and receiving elevation angle; (b) beam divergence angle and receiving field of view angle

Fig. 7. Loss curves for different initial learning rates

Fig. 8. Loss curves for different numbers of first layer LSTM neurons

Fig. 9. DNN hidden layer loss curves for different parameter values. (a) Number of hidden layers; (b) number of neurons in a single hidden layer

Fig. 10. Performance comparison of different channel equalization algorithms. (a) BER; (b) MSE