Channel Prediction Method and Related Device

Patent metadata

Field	Value
Publication number	US-2021194733-A1
Application number	US-202117196337-A
Country	US
Kind code	A1
Filing date	Mar 9, 2021
Priority date	Sep 10, 2018
Publication date	Jun 24, 2021
Grant date	—

Abstract

A channel prediction method and a related device, where the method includes: obtaining a first channel coefficient sequence in a first time period, where the first channel coefficient sequence includes a plurality of complex values of a channel coefficient; and determining a prediction value of the channel coefficient in a second time period based on the first channel coefficient sequence and a preset vocabulary of channel changes, where the preset vocabulary of channel changes includes a mapping relationship between a channel change value index and each change value of the channel coefficient, and where the second time period is later than the first time period.

First claim

What is claimed is:

1 . A channel prediction method, comprising:
obtaining a first channel coefficient sequence in a first time period, wherein the first channel coefficient sequence comprises a plurality of complex values of a channel coefficient; and determining a prediction value of the channel coefficient in a second time period based on the first channel coefficient sequence and a preset vocabulary of channel changes, wherein the preset vocabulary of channel changes comprises a mapping relationship between a channel change value index and each change value of the channel coefficient, and wherein the second time period is later than the first time period.

2 . The channel prediction method according to claim 1 , wherein determining the prediction value of the channel coefficient comprises:
determining a first channel coefficient change sequence based on the first channel coefficient sequence, wherein the first channel coefficient change sequence comprises a plurality of change values of the channel coefficient; searching the preset vocabulary of channel changes for a first channel change value index corresponding to each change value in the first channel coefficient change sequence; generating a first channel change value index sequence; inputting the first channel change value index sequence into a channel prediction model to obtain a channel change value index prediction sequence; and determining the prediction value of the channel coefficient based on the channel change value index prediction sequence.

3 . The channel prediction method according to claim 2 , wherein determining the first channel coefficient change sequence based on the first channel coefficient sequence comprises:
obtaining each change value in the first channel coefficient change sequence by subtracting a complex value of the channel coefficient at a former moment from a complex value of the channel coefficient at a latter moment in the first channel coefficient sequence; or obtaining each change value in the first channel coefficient change sequence by subtracting a complex value of the channel coefficient at a latter moment from a complex value of the channel coefficient at a former moment in the first channel coefficient sequence.

4 . The channel prediction method according to claim 2 , wherein determining the prediction value of the channel coefficient based on the channel change value index prediction sequence comprises:
determining, based on the preset vocabulary of channel changes, a channel coefficient change value prediction sequence corresponding to the channel change value index prediction sequence; and determining the prediction value of the channel coefficient based on the channel coefficient change value prediction sequence.

5 . The channel prediction method according to claim 1 , further comprising:
obtaining a second channel coefficient sequence, wherein the second channel coefficient sequence comprises a plurality of complex values of the channel coefficient; determining a second channel coefficient change sequence based on the second channel coefficient sequence, wherein the second channel coefficient change sequence comprises a plurality of change values of the channel coefficient; searching the preset vocabulary of channel changes for a first channel change value index corresponding to each change value in the second channel coefficient change sequence; generating a second channel change value index sequence; and obtaining a channel prediction model by inputting the second channel change value index sequence into a neural network for training.

6 . The channel prediction method according to claim 5 , further comprising:
obtaining a probability of each prediction value output by the neural network; determining a difference between complex values of every two change values in the preset vocabulary of channel changes; generating a channel change difference matrix; determining a weighted mean of probabilities of all prediction values based on the channel change difference matrix and the probability of each prediction value; and determining, based on the weighted mean, whether the neural network has completed training.

7 . The channel prediction method according to claim 5 , wherein an input dimension of the neural network and an output dimension of the neural network are equal to a quantity of mapping relationships in the preset vocabulary of channel changes.

8 . The channel prediction method according to claim 1 , further comprising:
collecting statistics on an occurrence frequency of each of a plurality of change values of the channel coefficient; obtaining the channel change value index by assigning an integer to each change value of the channel coefficient based on the occurrence frequency; and establishing the mapping relationship to generate the preset vocabulary of channel changes.

9 . The channel prediction method according to claim 8 , wherein obtaining the channel change value index further comprises assigning, when an occurrence frequency of a target change value in the plurality of change values is greater than a preset threshold, an integer to the target change value of the channel coefficient.

10 . A channel prediction apparatus, comprising:
a non-transitory memory configured to store instructions; and a processor coupled to the non-transitory memory and configured to execute the instructions to:
obtain a first channel coefficient sequence in a first time period, wherein the first channel coefficient sequence comprises a plurality of complex values of a channel coefficient; and
determine a prediction value of the channel coefficient in a second time period based on the first channel coefficient sequence and a preset vocabulary of channel changes,
wherein the preset vocabulary of channel changes comprises a mapping relationship between a channel change value index and each change value of the channel coefficient, and
wherein the second time period is later than the first time period.

11 . The channel prediction apparatus according to claim 10 , wherein the processor is further configured to execute the instructions to:
determine a first channel coefficient change sequence based on the first channel coefficient sequence, wherein the first channel coefficient change sequence comprises a plurality of change values of the channel coefficient; search the preset vocabulary of channel changes for a first channel change value index corresponding to each change value in the first channel coefficient change sequence; generate a first channel change value index sequence; input the first channel change value index sequence into a channel prediction model for prediction to obtain a channel change value index prediction sequence; and determine the prediction value of the channel coefficient based on the channel change value index prediction sequence.

12 . The channel prediction apparatus according to claim 11 , wherein the processor is further configured to execute the instructions to:
obtain each change value in the first channel coefficient change sequence by subtracting a complex value of the channel coefficient at a former moment from a complex value of the channel coefficient at a latter moment in the first channel coefficient sequence; or obtain each change value in the first channel coefficient change sequence by subtracting a complex value of the channel coefficient at a latter moment from a complex value of the channel coefficient at a former moment in the first channel coefficient sequence.

13 . The channel prediction apparatus according to claim 11 , wherein the processor is fu

Assignees

• Huawei Tech Co Ltd

Inventors

• Huangfu Yourui
• Li Rong
• Wang Jian
• Dai Shengchen
• Wang Jun
• Yu Rongdao

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/096

  Transfer learning · CPC title

• G06N3/09

  Supervised learning · CPC title