Fig 1.
Structure of the multi-leader self-attention mechanism.
Table 1.
Perception errors for different combinations of convolutional kernel sizes.
Table 2.
Perceived errors for different BiLSTM neuron numbers.
Fig 2.
Impact of MHSA heads on inference time and memory usage with distinct colors and markers.
Table 3.
Perceived errors for different MHSA parameters.
Fig 3.
Accuracy variation curve during model training.
Fig 4.
Loss variation curve during model training.
Table 4.
Information on different model configurations under the harmonized dataset.
Fig 5.
Spectrum utilization under different SU counts.
Fig 6.
Spectrum utilization under different transmission power levels.
Table 5.
Local feature extraction network ablative experimental perceptual error results.
Fig 7.
Comparison of perception errors in ablation study of the global feature extraction network.
Table 6.
Global feature extraction network ablativity experiment perception error results.
Table 7.
Detailed information of various deep learning models.
Fig 8.
Time consumption of different models.
Fig 9.
Comparison of results of different methods under various SU counts.
Table 8.
Comparative experimental results with different numbers of Sus.