Channel equalization device and method, and computer readable storage medium storing instructions to perform channel equalization method
There is provided a channel equalization device. The channel equalization device comprises a receiver configured to receive a plurality of consecutive reception symbol sequences through multiple channels; a memory storing one or more instructions; and a processor configured to execute the one or more instructions stored in the memory, wherein the instructions, when executed by the processor, cause the processor to estimate a transmission symbol at a position of interest among a plurality of consecutive transmission symbol sequences based on the plurality of received reception symbol sequences using a neural filter trained by training reception symbol sequences.
1 . A channel equalization device comprising:
a receiver configured to receive a plurality of consecutive reception symbol sequences through multiple channels of a wideband;
a memory storing one or more instructions; and
a processor configured to execute the one or more instructions stored in the memory, wherein the instructions, when executed by the processor, cause the processor to estimate a transmission symbol at a position of interest among a plurality of consecutive transmission symbol sequences based on the received plurality of consecutive reception symbol sequences using a neural filter to reduce a multi-channel tap effect of the received plurality of consecutive reception symbol sequences,
wherein the neural filter is trained by a plurality of training reception symbol sequences,
wherein the neural filter includes at least two 1-D convolution layers, and
wherein one of the at least two 1-D convolution layers is configured to expand channels of the received plurality of consecutive reception symbol sequences to generate a plurality of channels, and another of the at least two 1-D convolution layers is configured to reduce the generated plurality of channels into a predetermined number of channels.
2 . The channel equalization device of claim 1 , wherein the processor is configured to generate the neural filter by training the plurality of training reception symbol sequences.
3 . The channel equalization device of claim 2 , wherein the processor is configured to train the plurality of training reception symbol sequences labeled with a training transmission symbol at a learning position of interest among consecutive learning transmission symbol sequences.
4 . The channel equalization device of claim 2 , wherein the processor is configured to generate the neural filter using supervised learning.
5 . The channel equalization device of claim 1 , wherein the memory is configured to store the neural filter in advance and provide the neural filter stored in advance to the processor.
6 . The channel equalization device of claim 1 , further comprising a communication unit configured to receive the neural filter from the outside and provide the neural filter to the processor.
7 . The channel equalization device of claim 1 , wherein the processor is configured to transmit the received plurality of consecutive reception symbol sequences to a cloud including the neural filter and receive the transmission symbol at the position of interest output by the neural filter from the cloud.
8 . A channel equalization method comprising:
receiving a plurality of consecutive reception symbol sequences through multiple channels of a wideband; and
estimating a transmission symbol at a position of interest among a plurality of consecutive transmission symbol sequences based on the received plurality of consecutive reception symbol sequences using a neural filter to reduce a multi-channel tap effect of the received plurality of consecutive reception symbol sequences,
wherein the neural filter is trained by a plurality of training reception symbol sequences,
wherein the neural filter includes at least two 1-D convolution layers, and
wherein one of the at least two 1-D convolution layers is configured to expand channels of the received plurality of consecutive reception symbol sequences to generate a plurality of channels, and another of the at least two 1-D convolution layers is configured to reduce the generated plurality of channels into a predetermined number of channels.
9 . The channel equalization method of claim 8 , further comprising generating the neural filter by training the plurality of training reception symbol sequences.
10 . The channel equalization method of claim 9 , wherein the generating the neural filter includes learning the plurality of training reception symbol sequences labeled with a training transmission symbol at a learning position of interest among consecutive learning transmission symbol sequences.
11 . The channel equalization method of claim 9 , wherein the generating the neural filter includes generating the neural filter using supervised learning.
12 . The channel equalization method of claim 8 , further comprising pre-storing the neural filter in a memory,
wherein the estimating the transmission symbol at the position of interest includes estimating the transmission symbol at the position of interest using the pre-stored neural filter.
13 . The channel equalization method of claim 8 , wherein the estimating the transmission symbol at the position of interest includes receiving the neural filter from the outside.
14 . The channel equalization method of claim 13 , wherein the estimating the transmission symbol at the position of interest includes:
transmitting the received plurality of consecutive reception symbol sequences to a cloud including the neural filter; and
receiving the transmission symbol at the position of interest output by the neural filter from the cloud.
15 . A non-transitory computer readable storage medium storing computer executable instructions, wherein the instructions, when executed by a processor, cause the processor to perform a channel equalization method, the method comprising:
receiving a plurality of consecutive reception symbol sequences through multiple channels of a wideband; and
estimating a transmission symbol at a position of interest among a plurality of consecutive transmission symbol sequences based on the received plurality of consecutive reception symbol sequences using a neural filter to reduce a multi-channel tap effect of the received plurality of consecutive reception symbol sequences,
wherein the neural filter is trained by a plurality of training reception symbol sequences,
wherein the neural filter includes at least two 1-D convolution layers, and
wherein one of the at least two 1-D convolution layers is configured to expand channels of the received plurality of consecutive reception symbol sequences to generate a plurality of channels, and another of the at least two 1-D convolution layers is configured to reduce the generated plurality of channels into a predetermined number of channels.
16 . The non-transitory computer readable storage medium of claim 15 , further comprising generating the neural filter by training the plurality of training reception symbol sequences.
17 . The non-transitory computer readable storage medium of claim 16 , wherein the generating the neural filter includes learning the plurality of training reception symbol sequences labeled with a training transmission symbol at a learning position of interest among consecutive learning transmission symbol sequences.
18 . The non-transitory computer readable storage medium of claim 16 , wherein the generating the neural filter includes generating the neural filter using supervised learning.
19 . The non-transitory computer readable storage medium of claim 15 , further comprising pre-storing the neural filter in a memory,
wherein the estimating the transmission symbol at the position of interest includes estimating the transmission symbol at the position of interest using the pre-stored neural filter.
20 . The non-transitory computer readable storage medium of claim 15 , wherein the estimating the transmission symbol at the position of interest includes receiving the neural filter from the outside.