.png){kind=icon}
Return
Quantized Trainable Compressed Sensing for MIMO CSI Feedback
.png){kind=icon}
OA
AI
DOI：10.1109/TVT.2024.3446464.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Low-complexity CSI compression and feedback methods are essential for mobile communication systems, especially for resource-limited UEs. In this paper, a deep learning (DL)-based quantized trainable compressed sensing (QTCS) CSI feedback method is proposed. The encoder is very simple and only uses a single matrix-vector multiplication operation for realizing CSI compression, which greatly decreases the computation burden at the UE. The decoder module follows the iterative shrinkage-thresholding algorithm (ISTA) principle and the attention mechanism is used to recover the CSI. A vector quantize layer is introduced which enables the encoder and decoder to be jointly trained from end-to-end. Simulations demonstrate that the QTCS outperforms the existing methods in 3GPP UMi and UMa channels, even though the encoder is much simpler.
Keywords:
Quantization (signal)
Decoding
Sensors
Vectors
Training
Downlink
Estimation
Compressed sensing
trainable quantization
massive MIMO
CSI feedback
Journal
IF:
7.1
Papers:
1.8W
Citations:
6.6W
Organization
No organization information available
Cited Papers
.png){kind=icon}
ShareSaveShareSaveShareSaveShareSave
View more