US-12621068-B2 - Nonlinear neural network with phase normalization for base-band modelling of radio-frequency non-linearities

Abstract

Various example embodiments relate to mitigation of a non-linearity in a data communication chain. A method may include: obtaining a data communication signal including a plurality of complex-valued input samples; capturing, from the plurality of complex-valued input samples, a current sample and a set of delayed samples; generating a phase-normalized input signal based on normalizing phase of the current sample and the set of delayed samples by a normalization term, wherein the normalization term is common for the current sample and the set of delayed samples; providing the phase-normalized input signal to a neural network configured to mitigate non-linearity of a data communication chain and to output a complex-valued output sample for each of the plurality of complex-valued input samples; and denormalizing phase of the complex-valued output sample by a denormalization term configured to restore phase of the data communication signal.

Inventors

• Arne FISCHER-BUHNER

Assignees

• NOKIA SOLUTIONS AND NETWORKS OY

Dates

Publication Date

20260505

Application Date

20240327

Priority Date

20230404

Claims (14)

1. 1 . A method, comprising: obtaining a data communication signal comprising a plurality of complex-valued input samples; capturing, from the plurality of complex-valued input samples, a current sample and a set of delayed samples; generating a phase-normalized input signal based on normalizing phase of the current sample and the set of delayed samples with a normalization term, wherein the normalization term is common for the current sample and the set of delayed samples; providing the phase-normalized input signal to a neural network, the neural network configured to mitigate non-linearity of a data communication chain and to output a complex-valued output sample for the plurality of complex-valued input samples; and denormalizing phase of the complex-valued output sample with a denormalization term, the denormalization term configured to restore phase of the data communication signal; wherein the normalizing phase is based on multiplication of the current sample and the delayed samples with the normalization term, wherein the denormalizing phase is based on multiplication of the complex-valued output sample with the denormalization term, and wherein the denormalization term comprises a complex conjugate of the normalization term.
2. 2 . The method of claim 1 , wherein the neural network is a real-valued neural network.
3. 3 . The method of claim 1 , wherein the normalization term is configured to normalize current samples of the plurality of complex-valued input samples to zero-phase.
4. 4 . The method of claim 1 , wherein samples of the phase-normalized input signal are decomposed into two real-valued samples, wherein the two real-valued samples represent real and imaginary parts of the current sample or a delayed sample of the set of delayed samples.
5. 5 . The method of claim 1 , further comprising: providing, to the neural network, an augmented input comprising a second set of delayed samples comprising a p-th power of an absolute value of the set of delayed samples, where p is a non-zero integer.
6. 6 . The method of claim 1 , wherein the neural network is configured to mitigate non-linearity of a radio frequency part of the data communication chain.
7. 7 . The method of claim 1 , further comprising at least one of: providing the data communication signal to a linear dynamic model configured to mitigate the non-linearity of the data communication chain parallel to the neural network; or providing the data communication signal to a non-linear dynamic model configured to mitigate the non-linearity of the data communication parallel to the neural network.
8. 8 . The method of claim 1 , wherein the neural network is configured to mitigate a non-linearity of a radio frequency amplifier.
9. 9 . The method of claim 8 , wherein the radio frequency amplifier is a power amplifier.
10. 10 . The method of claim 1 , wherein the data communication signal comprises a baseband signal.
11. 11 . A non-transitory program storage device readable with an apparatus, tangibly embodying a program of instructions executable with the apparatus to perform the method according to claim 1 .
12. 12 . An apparatus, comprising: at least one processor; and at least one non-transitory memory configured to store instructions that, when executed with the at least one processor, cause the apparatus to perform: obtaining a data communication signal comprising a plurality of complex-valued input samples; capturing, from the plurality of complex-valued input samples, a current sample and a set of delayed samples; generating a phase-normalized input signal based on normalizing phase of the current sample and the set of delayed samples with a normalization term, wherein the normalization term is common for the current sample and the set of delayed samples; providing the phase-normalized input signal to a neural network, the neural network configured to mitigate non-linearity of a data communication chain and to output a complex-valued output sample for the plurality of complex-valued input samples; and denormalizing phase of the complex-valued output sample with a denormalization term, the denormalization term configured to restore phase of the data communication signal; wherein the normalizing phase is based on multiplication of the current sample and the delayed samples with the normalization term, wherein the denormalizing phase is based on multiplication of the complex-valued output sample with the denormalization term, and wherein the denormalization term comprises a complex conjugate of the normalization term.
13. 13 . The apparatus according to claim 12 , wherein the instructions when executed with the at least one processor, execute the method according to claim 2 .
14. 14 . The apparatus of claim 12 , wherein the apparatus comprises any of a transmitter, a receiver, a transceiver, a user equipment, or an access node.

Description

TECHNICAL FIELD Various example embodiments relate to mitigation of a non-linearity in a data communication chain. Some example embodiments relate to phase-normalization of an input signal for a neural network configured to mitigate the non-linearity. BACKGROUND Power-efficient transmission in any modern data communication chain is of key importance. Various circuit components may cause a number of non-linearity related problems. A non-linear digital model may be applied to mitigate such non-linearities in these circuit elements. A general requirement for these non-linear digital models is high modelling capability at low complexity and therefore neural networks (NN) may be considered for this purpose. SUMMARY This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Example embodiments of the present disclosure improve performance of mitigating non-linearity of a data communication chain. This and other benefits may be achieved by the features of the independent claims. Further example embodiments are provided in the dependent claims, the description, and the drawings. According to a first aspect, a method is disclosed. The method may comprise: obtaining a data communication signal comprising a plurality of complex-valued input samples; capturing, from the plurality of complex-valued input samples, a current sample and a set of delayed samples; generating a phase-normalized input signal based on normalizing phase of the current sample and the set of delayed samples by a normalization term, wherein the normalization term is common for the current sample and the set of delayed samples; providing the phase-normalized input signal to a neural network configured to mitigate non-linearity of a data communication chain and to output a complex-valued output sample for each of the plurality of complex-valued input samples; and denormalizing phase of the complex-valued output sample by a denormalization term configured to restore phase of the data communication signal. According to an example embodiment of the first aspect, the neural network is a real-valued neural network. According to an example embodiment of the first aspect, the normalization term may be configured to normalize each current sample of the plurality of complex-valued input samples to zero-phase. According to an example embodiment of the first aspect, each sample of the phase-normalized input signal is decomposed into two real-valued samples, wherein the two real-valued samples represent real and imaginary parts of the current sample or a delayed sample. According to an example embodiment of the first aspect, the normalization may be based on multiplication of the current sample and the delayed samples by the normalization term, and the denormalization may be based on multiplication of the complex-valued output sample by the denormalization term. The denormalization term may comprise a complex conjugate of the normalization term. According to an example embodiment of the first aspect, the method may further comprise providing, to the neural network an augmented input, comprising a second set of delayed samples comprising a p-th power of an absolute value of the delayed samples, where p is a non-zero integer. According to an example embodiment of the first aspect, the neural network may be further configured to mitigate non-linearity of a radio frequency part of the data communication chain. According to an example embodiment of the first aspect, the method may further comprise providing the data communication signal to a linear dynamic model configured to mitigate the non-linearity of the data communication chain parallel to the neural network. According to an example embodiment of the first aspect, the method may further comprise providing the data communication signal to a non-linear dynamic model configured to mitigate the non-linearity of the data communication chain parallel to the neural network. According to an example embodiment of the first aspect, the neural network may be further configured to mitigate a non-linearity of a radio frequency amplifier. According to an example embodiment of the first aspect, the radio frequency amplifier may comprise a power amplifier. According to an example embodiment of the first aspect, the data communication signal may comprise a baseband signal. According to an example embodiment of the first aspect, the method is performed by a transmitter, a receiver, a transceiver, a user equipment, or an access node. According to a second aspect, an apparatus is disclosed. The apparatus may comprise at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least