Research Article
A Convolutional Self-Attention Network for CSI Reconstruction in MIMO System
Table 4
NMSE (db) performance of the proposed quantization methods for different scenarios.
(a) |
| | Indoor | | CR | 4 | 16 | 32 | | B | 4 | 5 | 6 | 4 | 5 | 6 | 4 | 5 | 6 |
| | CSANet | -34.18 | -15.54 | -11.24 | | UQ | -23.59 | -27.16 | -30.34 | -11.39 | -13.22 | -14.79 | -9.11 | -10.43 | -10.93 | | NUQ | 25.36 | 28.23 | -31.97 | -12.65 | -14.23 | -15.03 | -9.58 | -10.64 | -11.04 |
|
|
(b) |
| | Outdoor | | CR | 4 | 16 | 32 | | B | 4 | 5 | 6 | 4 | 5 | 6 | 4 | 5 | 6 |
| | CSANet | -14.73 | -6.52 | -4.45 | | UQ | -12.37 | -12.66 | -13.18 | -5.43 | -5.87 | -6.02 | -2.93 | -3.41 | -3.62 | | UNQ | -13.24 | -13.42 | -14.23 | -5.94 | -6.21 | -6.33 | -3.49 | -4.01 | -4.31 |
|
|
