Data augmentation method and receiver

1 . A data augmentation method, wherein the method is performed by a receiver and comprises:

performing a data augmentation process for a result obtained by a first basic model of the receiver to obtain a first data augmented training set;

performing an online training fine tuning process for the first basic model based on the first data augmented training set to obtain a second basic model; and

stopping a loop in response to a loop stopping condition being satisfied;

wherein the performing a data augmentation process for a result obtained by a first basic model of the receiver to obtain a first data augmented training set comprises:

acquiring a first receiving signal;

inputting the first receiving signal into the first basic model of the receiver to obtain a first bitstream; and

performing the data augmentation process for the first bitstream to obtain a second bitstream;

wherein the performing an online training fine tuning process for the first basic model based on the first data augmented training set to obtain a second basic model comprises:

performing the online training fine tuning process for the first basic model based on a second bitstream to obtain the second basic model;

wherein the performing the data augmentation process for the first bitstream to obtain a second bitstream comprises:

selecting a target bitstream from the first bitstream, performing a binary process for the target bitstream, and obtaining a perturbation bit vector set; and

obtaining a second training set based on the perturbation bit vector set and a receiving signal set, wherein the receiving signal set is obtained based on the perturbation bit vector set;

wherein the performing the online training fine tuning process for the first basic model based on a second bitstream to obtain the second basic model comprises:

performing the online training fine tuning process for the first basic model based on the second training set to obtain the second basic model;

wherein the method further comprises:

inputting the first receiving signal into the second basic model to obtain a third bitstream;

wherein the stopping a loop in response to a loop stopping condition being satisfied comprises:

stopping the loop in response to a symbol error rate of the third bitstream being less than a preset symbol error rate threshold, and/or, the number of loops being equal to a preset number threshold.

2 . The method according to claim 1 , wherein after the performing an online training fine tuning process for the first basic model based on the first data augmented training set to obtain a second basic model and before the stopping a loop in response to a loop stopping condition being satisfied, the method further comprises:

performing the data augmentation process for a result obtained by the second basic model to obtain a second data augmented training set; and

performing the online training fine tuning process for the second basic model based on the first data augmented training set to obtain a third basic model.

3 . The method according to claim 1 , further comprising:

acquiring a channel set;

generating a source bitstream;

obtaining a receiving signal based on the channel set and the source bitstream;

obtaining a first training set based on the source bitstream and the receiving signal; and

pretraining the first training set to obtain the first basic model.

4 . The method according to claim 1 , wherein the performing a data augmentation process for a result obtained by a first basic model of the receiver to obtain a first data augmented training set comprises:

acquiring a second receiving signal;

inputting the second receiving signal into the first basic model to obtain a fourth bitstream;

performing a channel decoding process for the fourth bitstream to obtain a fifth bitstream;

performing the channel coding process for the fifth bitstream to obtain a recoded sixth bitstream; and

obtaining a third training set based on the sixth bitstream and the second receiving signal;

wherein the performing an online training fine tuning process for the first basic model based on the first data augmented training set to obtain a second basic model comprises:

performing the online training fine tuning process for the first basic model based on third training set in response to times of receiving signals satisfying a reception time threshold, to obtain the second basic model.

5 . The method according to claim 4 , further comprising:

acquiring a channel set;

generating a source bitstream;

performing a channel coding process for the source bitstream to obtain a coded bitstream;

obtaining a receiving signal based on the channel set and the coded bitstream;

obtaining a fourth training set based on the coded bitstream and the receiving signal; and

pretraining the fourth training set to obtain the first basic model.

6 . The method according to claim 1 , wherein the receiver comprises a terminal device or a network device.

7 . The method according to claim 6 , further comprising:

online training for a part of network layers in the terminal device in response to the receiver comprising the terminal device; and

online training for all or a part of network layers in the network device in response to the receiver comprising the network device.

8 . A receiver, comprising:

a memory, storing executable program codes; and

a processor, coupled to the memory;

wherein the processor is configured to:

perform a data augmentation process for a result obtained by a first basic model of the receiver to obtain a first data augmented training set;

perform an online training fine tuning process for the first basic model based on the first data augmented training set to obtain a second basic model; and

stop a loop in response to a loop stopping condition being satisfied;

wherein the performing a data augmentation process for a result obtained by a first basic model of the receiver to obtain a first data augmented training set comprises:

acquiring a first receiving signal;

inputting the first receiving signal into the first basic model of the receiver to obtain a first bitstream; and

performing the data augmentation process for the first bitstream to obtain a second bitstream;

wherein the performing an online training fine tuning process for the first basic model based on the first data augmented training set to obtain a second basic model comprises:

performing the online training fine tuning process for the first basic model based on a second bitstream to obtain the second basic model;

wherein the performing the data augmentation process for the first bitstream to obtain a second bitstream comprises:

selecting a target bitstream from the first bitstream, performing a binary process for the target bitstream, and obtaining a perturbation bit vector set; and

obtaining a second training set based on the perturbation bit vector set and a receiving signal set, wherein the receiving signal set is obtained based on the perturbation bit vector set;

wherein the performing the online training fine tuning process for the first basic model based on a second bitstream to obtain the second basic model comprises:

performing the online training fine tuning process for the first basic model based on the second training set to obtain the second basic model;

wherein the processor is further configured to:

input the first receiving signal into the second basic model to obtain a third bitstream; and

stop the loop in response to a symbol error rate of the third bitstream being less than a preset symbol error rate threshold, and/or, the number of loops being equal to a preset number threshold.

9 . The receiver according to claim 8 , wherein the processor is further configured to:

perform the data augmentation process for a result obtained by the second basic model to obtain a second data augmented training set; and

perform the online training fine tuning process for the second basic model based on the first data augmented training set to obtain a third basic model.

10 . The receiver according to claim 8 , wherein the processor is further configured to:

acquire a channel set;

generate a source bitstream;

obtain a receiving signal based on the channel set and the source bitstream;

obtain a first training set based on the source bitstream and the receiving signal; and

pretrain the first training set to obtain the first basic model.

11 . The receiver according to claim 8 , wherein the processor is further configured to:

acquire a second receiving signal;

input the second receiving signal into the first basic model to obtain a fourth bitstream;

perform a channel decoding process for the fourth bitstream to obtain a fifth bitstream;

perform the channel coding process for the fifth bitstream to obtain a recoded sixth bitstream;

obtain a third training set based on the sixth bitstream and the second receiving signal; and

perform the online training fine tuning process for the first basic model based on third training set in response to times of receiving signals satisfying a receiving time threshold, to obtain the second basic model.

12 . The receiver according to claim 11 , wherein the processor is further configured to:

acquire a channel set;

generate a source bitstream;

perform a channel coding process for the source bitstream to obtain a coded bitstream;

obtain a receiving signal based on the channel set and the coded bitstream;

obtain a fourth training set based on the coded bitstream and the receiving signal; and

pretrain the fourth training set to obtain the first basic model.

13 . The receiver according to claim 8 , comprising a terminal device or a network device.

14 . The receiver according to claim 13 , wherein processor is further configured to:

online train for a part of network layers in the terminal device in response to the receiver comprising the terminal device; and

online train for all or a part of network layers in the network device in response to the receiver comprising the network device.