Machine learning-based vector quantization method and device for limited feedback in wireless communication system
View Patent ↗The present disclosure relates to a machine learning-based vector quantization method and device for limited feedback in a wireless communication system. The method for providing feedback on a selection vector by a reception terminal in a wireless communication system according to one embodiment of the present disclosure comprises the steps of: obtaining a feature matrix on the basis of a reception signal from a transmission terminal; obtaining a selection vector with respect to one or more codeword candidates included in a predetermined codebook, on the basis of the feature matrix; and sending feedback on the selection vector to the transmission terminal, wherein the selection vector comprises the same number of elements as the number of the one or more codeword candidates and may comprise one or more elements which are not 0.
1 . A method of giving feedback on a selection vector by a receiver in a wireless communication system, the method comprising:
receiving, by the receiver, a signal from a transmitter;
obtaining, through a fully connected layer of a deep neural network at the receiver, a feature matrix based on the signal received from the transmitter;
based on the feature matrix, obtaining, by the receiver, the selection vector for at least one codeword candidate included in a predetermined codebook; and
transmitting, by the receiver, feedback information including the selection vector to the transmitter,
wherein the selection vector is configured with a same number of elements as the number of the at least one codeword candidate and includes at least one non-zero element,
wherein the selection vector specifies a weight-based combination for at least one of the at least one codeword candidate,
wherein the selection vector is obtained based on a normalized feature matrix, wherein the normalized feature matrix is obtained by applying normalization to the feature matrix, wherein e i*,l which is an l-th element of the selection vector e i* is expressed as in Equation below:
e
i
*
,
l
=
G
_
c
l
α
∑
j
=
1
2
B
G
_
c
j
α
where, G is the normalized feature matrix,
c x is a x-th codeword of the at least one codeword candidate,
2 B corresponds to the number of the at least one codeword candidate,
α is a predetermined positive real number.
2 . The method of claim 1 , wherein:
the normalization is row-wise normalization.
3 . The method of claim 1 , wherein:
the signal received from the transmitter includes channel information.
4 . A method of determining a precoder by a transmitter in a wireless communication system, the method comprising:
transmitting, by the transmitter, a signal to a receiver;
receiving, by the transmitter from the receiver, feedback information including at least one selection vector;
obtaining, by the transmitter, at least one codeword candidate by using the at least one selection vector;
obtaining, through a fully connected layer of a deep neural network at the transmitter, a feature matrix based on the at least one codeword candidate; and
determining, by the transmitter, the precoder based on the feature matrix,
wherein the at least one selection vector is configured with a same number of elements as the number of the at least one codeword candidate included in a predetermined codebook and includes at least one non-zero element,
wherein the at least one selection vector specifies a weight-based combination for at least one of the at least one codeword candidate,
wherein the at least one selection vector is obtained based on a normalized feature matrix, wherein the normalized feature matrix is obtained by applying normalization to the feature matrix,
wherein e i*,l which is an l-th element of the at least one selection vector e i* is expressed as in Equation below:
e
i
★
,
l
=
G
_
c
l
α
∑
j
=
1
2
B
G
_
c
j
α
where, G is the normalized feature matrix,
c x is a x-th codeword of the at least one codeword candidate,
2 B corresponds to the number of the at least one codeword candidate,
α is a predetermined positive real number.
5 . The method of claim 4 , wherein:
each of the at least one codeword candidates is obtained by a product of the at least one selection vector and the predetermined codebook.
6 . The method of claim 4 , wherein:
the precoder is obtained based on the normalized feature matrix.
7 . The method of claim 6 , wherein:
the normalization is row-wise normalization.
8 . A receiving device performing feedback in a wireless communication, the receiving device comprising:
a transceiver;
an antenna unit;
a memory; and
a processor,
wherein the processor is configured to:
receive, through the transceiver, a signal from a transmitting device,
obtain and store in the memory a feature matrix, through a fully connected layer of a deep neural network at the receiving device, based on the signal received from the transmitting device through the transceiver and stored in the memory;
based on the feature matrix, obtain and store in the memory a selection vector for at least one codeword candidate included in a predetermined codebook; and
transmit feedback information including the selection vector to the transmitting device through the transceiver,
wherein the selection vector is configured with a same number of elements as the number of the at least one codeword candidate and includes at least one non-zero element,
wherein the selection vector specifies a weight-based combination for at least one of the at least one codeword candidate,
wherein the selection vector is obtained based on a normalized feature matrix, wherein the normalized feature matrix is obtained by applying normalization to the feature matrix, wherein e i*,l which is an l-th element of the selection vector e i* is expressed as in Equation below:
e
i
★
,
l
=
G
_
c
l
α
∑
j
=
1
2
B
G
_
c
j
α
where, G is the normalized feature matrix,
c x is a x-th codeword of the at least one codeword candidate,
2 B corresponds to the number of the at least one codeword candidate,
α is a predetermined positive real number.
9 . A transmitting device determining a precoder in a wireless communication system, the transmitting device comprising:
a transceiver;
an antenna unit;
a memory; and
a processor,
wherein the processor is configured to:
transmit, through the transceiver, a signal to a receiving device;
receive and store in the memory feedback information including at least one selection vector from the receiving device through the transceiver;
obtain and store in the memory at least one codeword candidate by using the at least one section vector;
obtain and store in the memory, through a fully connected layer of a deep neural network at the transmitting device, a feature matrix based on the at least one codeword candidate; and
determine the precoder based on the feature matrix,
wherein the at least one selection vector is configured with a same number of elements as the number of the at least one codeword candidate included in a predetermined codebook and includes at least one non-zero element,
wherein the at least one selection vector specifies a weight-based combination for at least one of the at least one codeword candidate,
wherein the at least one selection vector is obtained based on a normalized feature matrix, wherein the normalized feature matrix is obtained by applying normalization to the feature matrix, wherein e i*,l which is an l-th element of the at least one selection vector e i* is expressed as in Equation below:
e
i
★
,
l
=
G
_
c
l
α
∑
j
=
1
2
B
G
_
c
j
α
where, G is the normalized feature matrix,
c x is a x-th codeword of the at least one codeword candidate,
2 B corresponds to the number of the at least one codeword candidate,
α is a predetermined positive real number.