Signal processing circuit, receiver, signal processing method, and program

Abstract

This makes it possible to mitigate performance degradation caused by mode-dependent losses without significantly increasing computational complexity. [Solution] The signal processing circuit includes a MIMO demodulation unit that includes a first MIMO filter and performs MIMO demodulation processing on multiple signals; a signal extraction unit that extracts a signal of a predetermined time slot from the output signal of the MIMO demodulation unit; a removal component generation unit that includes a second MIMO filter and generates a signal component to be removed, including interference components that occur between multiple modes, based on the signal output from the MIMO demodulation unit; an interference removal unit that generates a signal from which the signal component to be removed has been removed from the signal extracted by the signal extraction unit; and a coefficient control unit that uses the magnitude of the difference between the signal from which the signal component to be removed has been removed and a reference signal as a loss function, and adaptively controls the coefficients of the first MIMO filter and the second MIMO filter, respectively, using backpropagation and gradient descent.

Claims

1. A MIMO demodulation unit includes a first Multi-input Multi-output (MIMO) filter and performs MIMO demodulation processing on multiple signals corresponding to multiple modes, A signal extraction unit extracts signals for a predetermined time slot from the output signal output from the MIMO demodulation unit, A removal component generation unit, which includes a second MIMO filter, generates a removal component to be removed, including interference components generated between the multiple modes, based on the signal output from the MIMO demodulation unit. An interference removal unit generates a signal from which the signal component to be removed has been removed from the signal extracted by the signal extraction unit, A signal processing circuit comprising a coefficient control unit that uses the magnitude of the difference between the signal from which the target signal component has been removed and a reference signal as a loss function, and adaptively controls the coefficients of the first MIMO filter and the coefficients of the second MIMO filter, respectively, using backpropagation and gradient descent.

2. The signal processing circuit according to claim 1, wherein the signal extraction unit has a single-input single-output (SISO) filter for each of the plurality of modes that extracts a signal for a predetermined time slot from the output signal.

3. The signal processing circuit according to claim 1 or 2, wherein the coefficient control unit fixes the coefficient of the tap corresponding to the predetermined time slot in the diagonal component of the coefficient of the second MIMO filter to 0.

4. The removal component generation unit further includes a symbol determination unit that determines the symbol of the output signal output from the MIMO demodulation unit, The signal processing circuit according to claim 1 or 2, wherein the second MIMO filter generates the signal component to be removed based on the determination result of the symbol.

5. The symbol determination unit determines the symbol of the signal output from the MIMO demodulation unit using a soft determination process. The signal processing circuit according to claim 4, wherein the coefficient control unit controls the parameters in the soft determination process using the gradient of the loss function with respect to the output signal input to the symbol determination unit and the gradient of the loss function with respect to the parameters in the soft determination process.

6. The signal processing circuit according to claim 4, wherein the symbol determination unit determines the symbol of the signal output from the MIMO demodulation unit using hard determination processing.

7. The signal processing circuit according to claim 1 or 2, wherein the first MIMO filter compensates for the coupling and mode dispersion between modes occurring in the plurality of signals.

8. A signal processing circuit according to claim 1 or 2, A receiver comprising a coherent receiver that coherently receives the aforementioned plurality of signals.

9. Using the first Multi-input Multi-output (MIMO) filter, MIMO demodulation processing is performed on multiple signals corresponding to multiple modes. From the output signal of the MIMO demodulation process, a signal for a predetermined time slot is extracted. Using a second MIMO filter, a signal component to be removed, including interference components generated between the multiple modes, is generated based on the output signal that has undergone MIMO demodulation. A signal is generated by removing the signal component to be removed from the extracted signal. A signal processing method that uses the magnitude of the difference between the signal from which the target signal component has been removed and a reference signal as a loss function, and adaptively controls the coefficients of the first MIMO filter and the coefficients of the second MIMO filter, respectively, using backpropagation and gradient descent.

10. Using the first Multi-input Multi-output (MIMO) filter, MIMO demodulation processing is performed on multiple signals corresponding to multiple modes. From the output signal of the MIMO demodulation process, a signal for a predetermined time slot is extracted. Using a second MIMO filter, a signal component to be removed, including interference components generated between the multiple modes, is generated based on the output signal that has undergone MIMO demodulation. A signal is generated by removing the signal component to be removed from the extracted signal. A program that causes a processor to perform a process that adaptively controls the coefficients of the first MIMO filter and the coefficients of the second MIMO filter, respectively, using the magnitude of the difference between the signal from which the target signal component to be removed and the reference signal as the loss function, and employing backpropagation and gradient descent.

Images