Delta-sigma modulation apparatus, delta-sigma modulation method and recording medium

1 . A delta-sigma modulation apparatus comprising:

one or more memories storing instructions; and

one or more processors configured to execute the instructions to

perform delta-sigma modulation on a first signal as an input signal, and output a second signal;

learn a parameter for a transmission path distortion model including at least a mixture model and a filter structure, using the second signal and a third signal generated through a transmission path of the second signal; and

input the second signal to the transmission path distortion model, and output a fourth signal that is an approximated value of a signal which is generated through at least part of the transmission path of the second signal,

wherein the one or more processors are further configured to execute the instructions to perform the delta-sigma modulation on the first signal using the fourth signal, and output the second signal.

2 . The delta-sigma modulation apparatus according to claim 1 , wherein

the mixture model is configured to receive the second signal as an input signal and output a coefficient for the filter structure, and

the filter structure is configured to apply the coefficient to the second signal and output the fourth signal.

3 . The delta-sigma modulation apparatus according to claim 1 , wherein

the transmission path distortion model further includes a linear filter, and

the one or more processors are further configured to execute the instructions to apply the linear filter to an output of the filter structure and output the fourth signal.

4 . The delta-sigma modulation apparatus according to claim 3 , wherein the linear filter is finite impulse response (FIR) or infinite impulse response (IIR).

5 . The delta-sigma modulation apparatus according to claim 1 , wherein the mixture model includes a neural network including two or more layers.

6 . The delta-sigma modulation apparatus according to claim 5 , wherein the one or more processors are further configured to execute the instructions to

acquire information for reduction of part of operations in the mixture model and the filter structure, and

reduce the part of the operations, based on the information.

7 . The delta-sigma modulation apparatus according to claim 5 , wherein the one or more processors are further configured to execute the instructions to learn the parameter for the transmission path distortion model using a fifth signal and the third signal, the fifth signal being generated from the second signal through a model simulating the transmission path of the second signal.

8 . The delta-sigma modulation apparatus according to claim 7 , wherein the one or more processors are further configured to execute the instructions to determine, using error propagation learning, the parameter for the transmission path distortion model to minimize an error norm between the fifth signal and the third signal.

9 . A delta-sigma modulation method comprising:

performing delta-sigma modulation on a first signal as an input signal, and outputting a second signal;

learning a parameter for a transmission path distortion model including at least a mixture model and a filter structure, using the second signal and a third signal generated through a transmission path of the second signal;

inputting the second signal to the transmission path distortion model, and outputting a fourth signal that is an approximated value of a signal which is generated through at least part of the transmission path of the second signal; and

performing the delta-sigma modulation on the first signal using the fourth signal, and outputting the second signal.

10 . A non-transitory computer-readable recording medium having recorded thereon a program causing a processor to perform:

performing delta-sigma modulation on a first signal as an input signal, and outputting a second signal;

learning a parameter for a transmission path distortion model including at least a mixture model and a filter structure, using the second signal and a third signal generated through a transmission path of the second signal;

inputting the second signal to the transmission path distortion model, and outputting a fourth signal that is an approximated value of a signal which is generated through at least part of the transmission path of the second signal; and

performing the delta-sigma modulation on the first signal using the fourth signal, and outputting the second signal.