US-12626190-B2 - Method of analyzing wireless signals using multi-task learning-based spectral analysis learning model

US12626190B2US 12626190 B2US12626190 B2US 12626190B2US-12626190-B2

Abstract

A wireless signal spectral analysis method using a multi-task learning-based spectral analysis learning model, the wireless signal spectral analysis method may be provided. The analysis method according to an embodiment of the present disclosure may include: receiving a target signal of a target band, obtaining a training dataset through pre-processing of the target signal, performing wireless signal spectral analysis learning using the training dataset, and analyzing the target signal using a trained spectral analysis learning model, wherein the performing of wireless signal spectral analysis learning comprises: configuring task specific layers for respectively performing individual learning for a plurality of tasks to be analyzed and a shared layer for performing shared learning; learning, in the shared layer, correlation data that meets a predefined criterion in the training dataset; and individually learning, in each of the plurality of task specific layers, using an individual dataset required for each task in the training dataset and a result of learning the correlation data.

Inventors

Won Tae Kim
Han Jin Kim
Young Jin Kim

Assignees

Korea University Of Technology And Education Industry—University Cooperation Foundation

Dates

Publication Date: 20260512
Application Date: 20221207
Priority Date: 20220927

Claims (7)

1 . A wireless signal spectral analysis method using a multi-task learning-based spectral analysis learning model, the method comprising: receiving a target signal of a target band, obtaining a training dataset through pre-processing of the target signal, performing wireless signal spectral analysis learning using the training dataset, and analyzing the target signal using a trained spectral analysis learning model, wherein the performing of wireless signal spectral analysis learning comprises: configuring task specific layers for respectively performing individual learning for a plurality of tasks to be analyzed and a shared layer for performing shared learning; learning, in the shared layer, correlation data that meets a predefined criterion in the training dataset; and individually learning, in each of the plurality of task specific layers, using an individual dataset required for each task in the training dataset and a result of learning the correlation data, wherein the plurality of task specific layers comprise: a signal classification task layer configured to classify signals of the target signal, the signal classification task layer performing individual learning based on I/Q data and constellation data in a time domain of the target signal, and a task layer selected from the group consisting of: a channel classification task layer, the channel classification task layer performing individual learning based on spectral density data of the target signal and amplitude data in the time domain; and a power estimation task layer, the power estimation task layer performing individual learning based on at least one of spectral density data of the target signal, and amplitude and phase data in the time domain.
2 . The wireless signal spectral analysis method of claim 1 , wherein the training dataset is composed of I/Q signal data, and the performing of the wireless signal spectral analysis learning comprises: determining accuracy of the individual learning; obtaining a modified training dataset by performing Fast Fourier Transform (FFT) on the I/Q signal data when the accuracy is less than a predefined accuracy criterion; and performing the wireless signal spectral analysis learning using the modified training dataset.
3 . The wireless signal spectral analysis method of claim 2 , wherein the performing of the wireless signal spectral analysis learning comprises: changing the number of neurons in the shared layer to correspond to the modified training dataset; and performing learning using the changed shared layer and the plurality of task specific layers.
4 . The wireless signal spectral analysis method of claim 1 , wherein the task layer selected from the group consisting of the channel classification task layer and the power estimation task layer is the channel classification task layer.
5 . The wireless signal spectral analysis method of claim 4 , wherein the correlation data learned by the shared layer comprises a bandwidth feature of the target signal.
6 . The wireless signal spectral analysis method of claim 1 , wherein the task layer selected from the group consisting of the channel classification task layer and the power estimation task layer is the power estimation task layer.
7 . The wireless signal spectral analysis method of claim 6 , wherein the correlation data learned by the shared layer comprises at least one of a bandwidth feature, a channel feature, and an I/Q feature of the target signal.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the priority of the Korean Patent Applications NO 10-2022-0122762 filed on Sep. 27, 2022, in the Korean Intellectual Property Office, the disclosure of which is incorporated herein by reference. BACKGROUND 1. Field The present disclosure relates to a method of analyzing a spectrum of a wireless signal, and more particularly, to a method of learning spectral analysis of a wireless signal using multi-task learning, and analyzing a wireless signal using the spectral analysis. 2. Description of the Related Art With the recent rapid development of wireless communication and semiconductors, wireless communication technology has been applied not only to existing wireless communication devices such as smartphones and tablets, but also to home appliances, vehicles, and public facilities. Accordingly, the trend of coexistence of various wireless communication technologies within the same network and spectral band is accelerating. As the number of devices using various wireless communication technologies increases, the possibility that interference and influence between devices frequently occurs in a multi-dimensional spectral space rather than a one-dimensional space is increasing. Wireless communication technologies are different from each other in a modulation method, frequency band, band and bandwidth, maximum power, data rate, and transmission range used. For example, as shown in FIG. 1, different wireless communication technologies have different bands and bandwidths. In addition, as shown in FIGS. 2 to 6, different wireless communication technologies have different features (I/Q signal pattern, constellation pattern, amplitude/phase pattern, magnitude pattern in a frequency domain, etc.) according to modulation methods. A signal using a specific wireless communication technology leaves a footprint for the above features, and when the footprint of the signal is analyzed in an n-dimensional space, a path in which interference of wireless signals can be minimized and collision can be prevented may be selected. Existing spectrum sensing technologies (hole detection, energy detection, preamble detection, etc.) for preventing the interference and collision support signal classification in a one-dimensional space or merely classify signals, and there is a limit to efficient path determination through spectral analysis in a multi-dimensional space. DESCRIPTION OF EMBODIMENTS Technical Problem Provided are apparatuses and methods of analyzing a signal source in a multi-dimensional spectral space and performing learning on a footprint. Provided are apparatuses and methods capable of minimizing interference and collision of wireless communication signals by comprehensively learning and analyzing multi-dimensional spatial information. The technical problems of the present disclosure are not limited to the technical problems mentioned above, and other technical problems that are not mentioned can be clearly understood by one of ordinary skill in the art from the following description. Technical Solution As an embodiment of the present disclosure, a wireless signal spectral analysis method using a multi-task learning-based spectral analysis learning model, the wireless signal spectral analysis method may be provided. The method according to an embodiment of the present disclosure may include: receiving a target signal of a target band, obtaining a training dataset through pre-processing of the target signal, performing wireless signal spectral analysis learning using the training dataset, and analyzing the target signal using a trained spectral analysis learning model, wherein the performing of wireless signal spectral analysis learning comprises: configuring task specific layers for respectively performing individual learning for a plurality of tasks to be analyzed and a shared layer for performing shared learning; learning, in the shared layer, correlation data that meets a predefined criterion in the training dataset; and individually learning, in each of the plurality of task specific layers, using an individual dataset required for each task in the training dataset and a result of learning the correlation data. The training dataset according to an embodiment of the present disclosure may composed of I/Q signal data, and the performing of the wireless signal spectral analysis learning may comprise: determining accuracy of the individual learning; obtaining a modified training dataset by performing Fast Fourier Transform (FFT) on the I/Q signal data when the accuracy is less than a predefined accuracy criterion; and performing the wireless signal spectral analysis learning using the modified training dataset. The performing of the wireless signal spectral analysis learning according to an embodiment of the present disclosure may comprise: changing the number of neurons in the shared layer to correspond to the modified training dataset; and performing learning