Communication system and codec method based on deep learning and known channel state information

1. A communication system based on deep learning and known channel state information (CSI), comprising:

a first electronic apparatus including a known first link CSI and a CSI encoder having a deep learning function; and

a second electronic apparatus including a known second link CSI and a CSI decoder having another deep learning function,

wherein the first link CSI and the second link CSI have a correlation or a similarity, the CSI encoder of the first electronic apparatus encodes or compresses the first link CSI into a first codeword, and feeds the first codeword back to the second electronic apparatus via a feedback link, and

wherein the CSI decoder of the second electronic apparatus encodes or compresses the second link CSI of the second electronic apparatus into a second codeword, and decodes or restores the first link CSI of the first electronic apparatus based on the first codeword encoded or compressed by the CSI encoder of the first electronic apparatus and the second codeword encoded or compressed by the CSI decoder of the second electronic apparatus.

2. The communication system of claim 1 , wherein the CSI encoder of the first electronic apparatus pre-processes the first link CSI, and includes an encoder configured for encoding or compressing the pre-processed first link CSI into the first codeword.

3. The communication system of claim 2 , wherein the encoder of the first electronic apparatus comprises at least one of a convolution layer, a batch normalization layer, an activation function layer and a pooling layer, provides an encoding process or a compressing process using the at least one of the convolution layer, the batch normalization layer, the activation function layer and the pooling layer, and encodes or compresses the first link CSI into the first codeword based on the encoding process or the compressing process.

4. The communication system of claim 1 , wherein the CSI decoder of the second electronic apparatus comprises:

an encoder configured for encoding or compressing the second link CSI into the second codeword; and

a decoder configured for decoding or restoring the first codeword and the second codeword into a three-layer dimension channel matrix.

5. The communication system of claim 4 , wherein the encoder of the second electronic apparatus comprises at least one of a convolution layer, a batch normalization layer, an activation function layer and a pooling layer, provides an encoding process and a compressing process using the at least one of the convolution layer, the batch normalization layer, the activation function layer and the pooling layer, and encodes or compresses the second link CSI into the second codeword based on the encoding process or the compressing process.

6. The communication system of claim 4 , wherein the decoder of the second electronic apparatus comprises at least one of a bilinear upsampling layer, a convolution layer, a batch normalization layer, a first activation function layer and a second activation function layer, provides a decoding process or a restoring process using the at least one of the bilinear upsampling layer, the convolution layer, the batch normalization layer, the first activation function layer and the second activation function layer, and decodes or restores the first codeword and the second codeword into the three-layer dimension channel matrix based on the decoding process or the restoring process.

7. The communication system of claim 4 , wherein the CSI decoder of the second electronic apparatus further comprises a neutral network having a memorizing function configured for decoding or restoring the first link CSI from the three-layer dimension channel matrix.

8. The communication system of claim 7 , wherein the neutral network having the memorizing function is a recurrent neutral network (RNN) or a long short-term memory (LSTM) network composed of multiple layers of LSTM units.

9. A codec method based on deep learning and known channel state information (CSI), comprising:

providing a first electronic apparatus including a known first link CSI and a CSI encoder having a deep learning function and a second electronic apparatus including a known second link CSI and a CSI decoder having another deep learning function, wherein the first link CSI and the second link CSI have a correlation or a similarity;

encoding or compressing, by the CSI encoder of the first electronic apparatus, the first link CSI into a first codeword, feeding the first codeword back to the second electronic apparatus via a feedback link, and encoding or compressing, by the CSI decoder of the second electronic apparatus, the second link CSI of the second electronic apparatus into a second codeword; and

decoding or restoring, by the CSI decoder of the second electronic apparatus, the first link CSI of the first electronic apparatus based on the first codeword encoded or compressed by the CSI encoder of the first electronic apparatus and the second codeword encoded or compressed by the CSI decoder of the second electronic apparatus.

10. The codec method of claim 9 , further comprising pre-processing, by the CSI encoder of the first electronic apparatus, the first link CSI, wherein the CSI encoder of the first electronic apparatus includes an encoder configured for encoding or compressing the pre-processed first link CSI into the first codeword.

11. The codec method of claim 10 , further comprising providing an encoding process or a compressing process using at least one of a convolution layer, a batch normalization layer, an activation function layer and a pooling layer of the encoder and encoding or compressing the first link CSI into the first codeword based on the encoding process or the compressing process.

12. The codec method of claim 9 , further comprising encoding or compressing, by an encoder of the CSI decoder of the second electronic apparatus, the second link CSI into the second codeword, and decoding or restoring, by a decoder of the CSI decoder of the second electronic apparatus, the first codeword and the second codeword into a three-layer dimension channel matrix.

13. The codec method of claim 12 , further comprising providing an encoding process or a compressing process using at least one of a convolution layer, a batch normalization layer, an activation function layer and a pooling layer of the encoder and encoding or compressing the second link CSI into the second codeword based on the encoding process or the compressing process.

14. The codec method of claim 12 , further comprising providing a decoding process or a restoring process using at least one of a bilinear upsampling layer, a convolution layer, a batch normalization layer, a first activation function layer and a second activation function layer of the decoder and decoding or restoring the first codeword and the second codeword into the three-layer dimension channel matrix based on the decoding process or the restoring process.

15. The codec method of claim 12 , further comprising restoring, by a neutral network having a memorizing function in the CSI decoder of the second electronic apparatus, the first link CSI from the three-layer dimension channel matrix.

16. The codec method of claim 15 , wherein the neutral network having the memorizing function is a recurrent neutral network (RNN) or a long short-term memory (LSTM) network composed of multiple layers of LSTM units.