Neural network-based channel estimation method and communication apparatus
This application discloses example neural network-based channel estimation methods and communication apparatuses. One example method may include obtaining first location information of a first communication apparatus, and processing the first location information by using a neural network to obtain first channel information. The first channel information is information about a radio channel between the first communication apparatus and a second communication apparatus and performing communication based on the first channel information. The neural network is used to estimate channel information based on location information.
1 . A neural network-based channel estimation method, wherein the method comprises:
obtaining first location information of a first communication apparatus;
processing the first location information by using a neural network to obtain first channel information, wherein the first channel information is information about a radio channel between the first communication apparatus and a second communication apparatus, and wherein processing the first location information by using the neural network to obtain the first channel information comprises:
processing the first location information to obtain second channel information, wherein a dimension of the second channel information is lower than a dimension of the first channel information; and
processing the second channel information to obtain the first channel information; and
performing communication based on the first channel information.
2 . The method according to claim 1 , wherein a parameter of the neural network is obtained through training based on historical data, the historical data comprises one or more pieces of mapping between historical location information and historical channel information, the historical location information is location information of a training apparatus during communication with the second communication apparatus, and the historical channel information is channel information of the training apparatus during communication with the second communication apparatus.
3 . The method according to claim 2 , wherein the method further comprises:
performing first training on the neural network based on the historical channel information, wherein a process of first training comprises:
changing a dimension of the historical channel information from a first dimension to a second dimension and changing the dimension of the historical channel information from the second dimension to the first dimension; and
performing second training on the neural network based on the historical location information corresponding to the historical channel information and the historical channel information in the second dimension.
4 . The method according to claim 1 , wherein an activation function of the neural network is a periodic function.
5 . The method according to claim 1 , wherein the neural network comprises:
Φ
(
x
)
=
W
n
(
ϕ
n
-
1
·
ϕ
n
-
2
·
…
·
ϕ
0
)
(
x
)
+
b
n
,
wherein
an input for an i th layer of the neural network is x i , a dimension of the input for the i th layer of the neural network is M i (M i >0), and i=0, 1, . . . , n−1;
an output of the i th layer of the neural network satisfies φ i (x i ), φ i (x i )=sin (W i x i +b i ), and a dimension of the output of the i th layer of the neural network is N i (N i >0);
a weight of the neural network satisfies W i ∈R N i ×M i ; a bias of the neural network satisfies b i ∈R N i ;
a sine function sin is used as a nonlinear activation function of the neural network;
x is an input for the neural network; and
Φ(x) is an output of the neural network.
6 . The method according to claim 1 , wherein the processing the first location information by using a neural network to obtain first channel information comprises:
processing the first location information to obtain one or more pieces of second location information, wherein the second location information is a function for a mirror point of the first location information; and
processing the one or more pieces of second location information to obtain the first channel information.
7 . The method according to claim 6 , wherein a dimension of the second location information is the same as a dimension of the first location information.
8 . The method according to claim 6 , wherein the neural network comprises an intermediate layer, and a quantity of neurons at the intermediate layer is an integer multiple of a dimension of the first location information.
9 . The method according to claim 8 , wherein the neural network further comprises a radial basis function (RBF) layer configured to process an output of the intermediate layer.
10 . The method according to claim 9 , wherein an activation function of the RBF layer is a periodic kernel function.
11 . The method according to claim 9 , wherein an activation function of the RBF layer satisfies the following formula:
ϕ
(
x
)
=
cos
(
w
❘
"\[LeftBracketingBar]"
x
-
a
❘
"\[RightBracketingBar]"
+
b
)
exp
(
-
❘
"\[LeftBracketingBar]"
x
-
c
❘
"\[RightBracketingBar]"
2
2
σ
2
)
,
wherein
x is an input for the RBF layer, φ(x) is an output of the RBF layer, and a, b, c, w, and σ are to-be-trained parameters.
12 . The method according to claim 1 , wherein the first channel information comprises uplink channel information and/or downlink channel information.
13 . A communication apparatus, comprising at least one processor and a communication interface, wherein the communication interface is configured to communicate with another apparatus, and the at least one processor is configured to execute a group of programs, and, when the groups of programs are executed by the processor, the communication apparatus is caused to:
obtain first location information of a first communication apparatus;
process the first location information by using a neural network to obtain first channel information, wherein the first channel information is information about a radio channel between the first communication apparatus and a second communication apparatus, and wherein processing the first location information by using the neural network to obtain the first channel information comprises:
processing the first location information to obtain second channel information, wherein a dimension of the second channel information is lower than a dimension of the first channel information; and
processing the second channel information to obtain the first channel information; and
perform communication based on the first channel information.
14 . The communication apparatus according to claim 13 , wherein a parameter of the neural network is obtained through training based on historical data, the historical data comprises one or more pieces of mapping between historical location information and historical channel information, the historical location information is location information of a training apparatus during communication with the second communication apparatus, and the historical channel information is channel information of the training apparatus during communication with the second communication apparatus.
15 . The communication apparatus according to claim 14 , wherein when the groups of programs are executed by the at least one processor, the communication apparatus is caused to:
perform first training on the neural network based on the historical channel information, wherein a process of first training comprises: changing a dimension of the historical channel information from a first dimension to a second dimension and changing the dimension of the historical channel information from the second dimension to the first dimension; and
perform second training on the neural network based on the historical location information corresponding to the historical channel information and the historical channel information in the second dimension.
16 . The communication apparatus according to claim 13 , wherein an activation function of the neural network is a periodic function.
17 . The communication apparatus according to claim 13 , wherein the neural network comprises:
Φ
(
x
)
=
W
n
(
ϕ
n
-
1
·
ϕ
n
-
2
·
…
·
ϕ
0
)
(
x
)
+
b
n
,
wherein
an input for an i th layer of the neural network is x i , a dimension of the input for the i th layer of the neural network is M i (M i >0), and i=0, 1, . . . , n−1;
an output of the i th layer of the neural network satisfies φ i (x i ), φ i (x i )=sin (W i x i +b i ), and a dimension of the output of the i th layer of the neural network is N i (N i >0);
a weight of the neural network satisfies W i ∈R N i ×M i ;
a bias of the neural network satisfies b i ∈R N i ;
a sine function sin is used as a nonlinear activation function of the neural network;
x is an input for the neural network; and
Φ(x) is an output of the neural network.
18 . The communication apparatus according to claim 13 , wherein when the groups of programs are executed by the at least one processor, the communication apparatus is caused to:
process the first location information to obtain one or more pieces of second location information, wherein the second location information is a function for a mirror point of the first location information; and
process the one or more pieces of second location information to obtain the first channel information.
19 . The method according to claim 1 , wherein the first communication apparatus is a terminal device, and the second communication apparatus is a network device.
20 . The communication apparatus according to claim 13 , wherein the first communication apparatus is a terminal device, and the second communication apparatus is a network device.