Method and device for decoding a signal, and memory device

Patent metadata

Field	Value
Publication number	US-10924304-B2
Application number	US-201716062486-A
Country	US
Kind code	B2
Filing date	Aug 28, 2017
Priority date	Aug 28, 2017
Publication date	Feb 16, 2021
Grant date	Feb 16, 2021

Abstract

A method for decoding a signal. The method comprises decoding a received signal according to a predetermined decoding scheme and obtaining an estimate of a transmission signal based on the decoding result; obtaining an estimate of channel noise by using the correlation of channel noise; generating a modified received signal by subtracting the obtained estimate of channel noise from the received signal; and decoding the modified received signal to obtain a decoded signal according to the predetermined decoding scheme.

First claim

We claim:

1. A method for decoding a signal, comprising:
decoding a received signal according to a predetermined decoding scheme and obtaining an estimate of a transmission signal based on the decoding result;
obtaining an estimate of channel noise by using the correlation of channel noise;
generating a modified received signal by subtracting the obtained estimate of channel noise from the received signal; and
decoding the modified received signal to obtain a decoded signal according to the predetermined decoding scheme
wherein the obtaining an estimate of channel noise by using the correlation of channel noise comprises:
obtaining a first estimate of channel noise by subtracting the estimate of the transmission signal from the received signal; and
processing the obtained first estimate of channel noise to obtain a second estimate of channel noise as the estimate of channel noise by using the correlation of channel noise.

2. The method of claim 1 , wherein the processing the obtained first estimate of the channel noise to obtain a second estimate of the channel noise as the estimate of the channel noise by using the correlation of channel noise comprises:
taking the first estimate of channel noise as an input to a convolutional neural network and taking an output of the convolutional neural network as the second estimate of the channel noise.

3. The method of claim 2 , further comprising:
training the convolution neural network by using a predetermined strategy, wherein the predetermined strategy includes any one of a training strategy that minimizes residual noise power, and a training strategy that causes distribution of residual noise to approximate to a Gaussian distribution while minimizing residual noise power.

4. The method of claim 1 , further comprising:
iteratively performing operations of decoding the received signal, obtaining an estimate of channel noise based on the correlation of channel noise, and generating the modified received signal by taking the modified received signal as a received signal, until a system state meets predetermined requirement.

5. A device for decoding a signal, comprising:
a decoder configured to decode a received signal according to a predetermined decoding scheme;
a transmission signal estimator configured to obtain an estimate of a transmission signal based on the decoding result;
a noise estimator configured to obtain an estimate of channel noise by using the correlation of channel noise; and
a modification signal generator configured to generate a modified received signal by subtracting the obtained estimate of channel noise from the received signal;
wherein the decoder is further configured to decode the modified received signal to obtain a decoded signal according to the predetermined decoding scheme,
wherein the noise estimator comprises:
a first noise estimation module configured to obtain a first estimate of channel noise by subtracting the estimate of the transmission signal from the received signal; and
a second noise estimation module configured to process the obtained first estimate of the channel noise to obtain a second estimate of the channel noise as the estimate of the channel noise by using the correlation of channel noise.

6. The device of claim 5 , wherein the second noise estimation module is implemented by a convolutional neural network, and wherein the second noise estimation module is further configured to:
take the first estimate of channel noise as an input to the convolutional neural network and take an output of the convolutional neural network as the second estimate of the channel noise.

7. The device of claim 6 , further comprising:
a network trainer configured to train the convolution neural network by using a predetermined strategy, wherein the predetermined strategy includes any one of a training strategy that minimizes residual noise power, and a training strategy that causes distribution of residual noise close to Gaussian distribution while minimizing residual noise power.

8. The device of claim 5 , wherein operations of the decoder, the transmission signal estimator, the noise estimator and the modification signal generator are iteratively performed by taking the modified received signal as a received signal, until a system state meets predetermined requirement.

9. A non-volatile, non-transitory storage medium on which a computer program is stored, which computer program, when executed by a processor, causes the processor to carry out the method according to claim 1 .

Assignees

• Univ Science & Technology China

Inventors

• Shen Cong
• Liang Fei
• Wu Feng

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N3/09

  Supervised learning · CPC title

• H04L25/0224 Primary

  using sounding signals · CPC title