Device for estimating channel in wireless communication system
View Patent ↗A method and a device for estimating a channel in a wireless communication system are provided. Provided is a meta learning device including processing circuitry configured to receive pilot signals from a plurality of first user equipment (UEs) to obtain received pilot signals, the received pilot signals being based on existing wireless communication connections of the plurality of first UEs, and estimate channels of a plurality of second UEs via meta learning using the received pilot signals, the channels corresponding to new wireless communication connections of the plurality of second UEs.
1 . A meta learning device comprising:
processing circuitry configured to
receive pilot signals from a plurality of first user equipment (UEs) to obtain received pilot signals, the received pilot signals being based on existing wireless communication connections of the plurality of first UEs, and
estimate channels of a plurality of second UEs via meta learning using the received pilot signals, the channels corresponding to new wireless communication connections of the plurality of second UEs.
2 . The meta learning device of claim 1 , wherein the processing circuitry is configured to estimate the channels using Model Agnostic Meta-Learning (MAML).
3 . The meta learning device of claim 1 , wherein the processing circuitry is configured to perform the meta learning by executing a meta training stage, a meta adaptation stage, and a meta testing stage.
4 . The meta learning device of claim 3 , wherein the processing circuitry is configured to:
first define a first least square channel estimate as a source data set of the meta training stage, the first least square channel estimate being obtained from a first UE among the plurality of first UEs; and
perform the meta learning based on the first least square channel estimate.
5 . The meta learning device of claim 4 , wherein the processing circuitry is configured to:
second define a plurality of second least square channel estimates as a target data set of each of the meta adaptation stage and the meta testing stage, each respective second least square channel estimate among the plurality of second least square channel estimates being obtained from a different one among the plurality of second UEs; and
perform the meta learning based on the plurality of second least square channel estimates.
6 . The meta learning device of claim 3 , wherein the processing circuitry is configured to define training data as a support set and define validation data as a query set in the meta training stage,
wherein an intersection between the support set and the query set is an empty set.
7 . The meta learning device of claim 3 , wherein the processing circuitry is configured to:
define a first data set as an adaptation set in the meta adaptation stage; and
define a second data set as a testing set in the meta testing stage,
wherein an intersection between the adaptation set and the testing set is an empty set.
8 . The meta learning device of claim 1 , wherein the processing circuitry is configured to perform the meta learning to train a neural network model.
9 . A base station of a wireless communication system, a plurality of first user equipment (UEs) having been connected to the base station, and the base station comprising:
processing circuitry configured to
receive pilot signals from the plurality of first UEs to obtain received pilot signals, and
estimate channels of a plurality of second UEs via meta learning using the received pilot signals, the plurality of second UEs being newly connected to the base station.
10 . The base station of claim 9 , wherein the processing circuitry is configured to estimate the channels via meta learning including estimating the channels using Model Agnostic Meta-Learning (MAML).
11 . The base station of claim 9 , wherein the processing circuitry is configured to perform the meta learning by executing a meta training stage, a meta adaptation stage, and a meta testing stage.
12 . The base station of claim 11 , wherein the processing circuitry is configured to:
first define a first least square channel estimate as a source data set of the meta training stage, the first least square channel estimate being obtained by the base station from a first UE among the plurality of first UEs;
second define a plurality of second least square channel estimates as a target data set of each of the meta adaptation stage and the meta testing stage, each respective second least square channel estimate among the plurality of second least square channel estimates being obtained by the base station from a different one among the plurality of second UEs; and
perform the meta learning based on the first least square channel estimate and the plurality of second least square channel estimates.
13 . The base station of claim 11 , wherein the processing circuitry is configured to define training data as a support set and define validation data as a query set in the meta training stage,
wherein an intersection between the support set and the query set is an empty set.
14 . The base station of claim 11 , wherein the processing circuitry is configured to:
define a first data set as an adaptation set in the meta adaptation stage; and
define a second data set as a testing set in the meta testing stage,
wherein an intersection between the adaptation set and the testing set is an empty set.
15 . The base station of claim 9 , wherein the processing circuitry is configured to perform the meta learning to train a neural network model.
16 . The base station of claim 9 , wherein the processing circuitry is configured to remove noise contained in an input value of the meta learning.
17 . The base station of claim 16 , wherein the processing circuitry is configured to remove the noise contained in the input value of the meta learning using a Deep Image Prior (DIP) scheme.
18 . A channel estimating device of a wireless communication system, the channel estimating device comprising:
processing circuitry configured to
receive pilot signals from a plurality of first user equipment (UEs) to obtain received pilot signals, the plurality of first UEs having been connected to a base station of the wireless communication system,
estimate channels of a plurality of second UEs via meta learning using the received pilot signals, the plurality of second UEs being newly connected to the base station, and
remove noise contained in an input value of the meta learning.
19 . The channel estimating device of claim 18 , wherein the processing circuitry is configured to estimate the channels using Model Agnostic Meta-Learning (MAML).
20 . The channel estimating device of claim 18 , wherein the processing circuitry is configured to perform the meta learning by executing a meta training stage, a meta adaptation stage, and a meta testing stage.