Method and apparatus for transceiving and receiving wireless signal in wireless communication system
According to the present document, a method by which a terminal receives data in a wireless communication system comprises: receiving a channel signal and a reference signal (RS) from a base station; generating a sequence by performing an operation of equalizing the RS to a channel RS; and decoding the received channel signal on the basis of the generated sequence, wherein the operation of equalizing the RS to the channel RS is based on a parameter determined according to a machine learning process.
1. A method performed by an apparatus, the method comprising:
training parameters of a deep neural network (DNN) based MIMO detector by using a training sequence set {H ref x+n} for x∈X as input, a binary sequence subset {b label } for b label ∈B as label, and a cross entropy function as a cost function, where H ref is a standard channel, x is a modulation symbol, X is a modulation symbol set, B is a binary sequence set, and n is additive white Gaussian noise;
training parameters of a DNN based preprocessor to obtain a fixed DNN based preprocessor by using a training sequence set {Hx+n} for x∈X as input, and s label =H ref x, which is a noiseless signal having undergone H ref , as label, where H is a current channel; and
fine-tuning the parameters of the DNN based MIMO detector by using output of the fixed DNN based preprocessor as a training input set;
receiving a first channel signal r;
obtaining a second channel signal r* based on the first channel signal r through the fixed DNN based preprocessor trained to transform a signal that has undergone an actual channel into a signal that undergoes the standard channel; and
obtaining a transmitted channel signal from the second channel signal r* through the DNN based MIMO detector trained for the standard channel.
2. An apparatus, the apparatus comprising:
at least one processor; and
at least one computer memory operably connectable to the at least one processor and storing instructions that, when executed, cause the at least one processor to perform operations comprising:
training parameters of a deep neural network (DNN) based MIMO detector by using a training sequence set {H ref x+n} for x∈X as input, a binary sequence subset {b label } for b label ∈B as label, and a cross entropy function as a cost function, where H ref is a standard channel, x is a modulation symbol, X is a modulation symbol set, B is a binary sequence set, and n is additive white Gaussian noise;
training parameters of a DNN based preprocessor to obtain a fixed DNN based preprocessor by using a training sequence set {Hx+n} for x∈X as input, and s label =H ref x, which is a noiseless signal having undergone H ref , as label, where H is a current channel; and
fine-tuning the parameters of the DNN based MIMO detector by using output of the fixed DNN based preprocessor as a training input set;
receiving a first channel signal r;
obtaining a second channel signal r* based on based on the first channel signal r through the fixed DNN based preprocessor trained to transform a signal that has undergone an actual channel into a signal that undergoes the standard channel; and
obtaining a transmitted channel signal from the second channel signal r* through the DNN based MIMO detector trained for the standard channel.
3. A non-transitory computer readable storage medium storing at least one computer program including instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising:
training parameters of a deep neural network (DNN) based MIMO detector by using a training sequence set {H ref x+n} for x∈X as input, a binary sequence subset {b label } for b label ∈B as label, and a cross entropy function as a cost function, where H ref is a standard channel, x is a modulation symbol, X is a modulation symbol set, B is a binary sequence set, and n is additive white Gaussian noise;
training parameters of a DNN based preprocessor to obtain a fixed DNN based preprocessor by using a training sequence set {Hx+n} for x∈X as input, and s label =H ref x, which is a noiseless signal having undergone H ref , as label, where H is a current channel; and
fine-tuning the parameters of the DNN based MIMO detector by using output of the fixed DNN based preprocessor as a training input set;
receiving a first channel signal r;
obtaining a second channel signal r* through the fixed DNN based preprocessor trained to transform a signal that has undergone an actual channel into a signal that undergoes the standard channel; and
obtaining a transmitted channel signal from the second channel signal r* through the DNN based MIMO detector trained for the standard channel.