| CPC H04B 7/043 (2013.01) [H04B 7/0456 (2013.01)] | 5 Claims |

|
1. A method for detecting transmitted data by a deep neural network, the method comprising:
receiving a plurality of complex data vectors (Y) and a plurality of MIMO-channel-transfer-matrixes-H, each complex data vector (Y) of said plurality of complex data vectors (Y) being associated with a MIMO-channel-transfer-matrix-H of said plurality of MIMO-channel-transfer-matrixes-H, each said MIMO-channel-transfer-matrix-H being associated with a certain Multiple Input Multiple Output (MIMO) channel;
decomposing each of said MIMO-channel-transfer-matrixes-H into a Q matrix and an R matrix;
transposing said Q matrix into QT;
multiplying each of said QT by an associated vector (Y) to thereby providing QT*Y;
training said deep neural network with a training set labeled with Log-Likelihood Ratio values; said training set comprising said plurality of said QT*Y and an upper part of said R matrix; and
utilizing said deep neural network for receiving second Log-Likelihood Ratio values from a second complex data vector (Y) and a second associated MIMO-channel-transfer-matrix-H and for detecting X from said second Log-Likelihood Ratio values; said X being a detected data vector detecting said transmitted data.
|