CN-122027411-A - Modulated signal recognition method, apparatus, device, medium, and program product

CN122027411ACN 122027411 ACN122027411 ACN 122027411ACN-122027411-A

Abstract

The application discloses a modulation signal identification method, a device, equipment, a medium and a program product, wherein the modulation signal identification method comprises the steps of acquiring a first signal of a modulation type to be identified; the first signal is input into a signal type recognition model to obtain the modulation type of the first signal output by the signal type recognition model, wherein the signal type recognition model is used for pre-training an initial model for carrying out modulation type recognition on the signal according to the received signal and the modulation type of the signal output signal, and the initial model is subjected to migration training. The application improves the accuracy of the modulation type identification of the signal.

Inventors

PENG YIXUAN
ZHENG KANG
HUA LEI
Qian Yuyang
SI YAXIONG

Assignees

中国移动紫金(江苏)创新研究院有限公司
中国移动通信集团江苏有限公司
中国移动通信集团有限公司

Dates

Publication Date: 20260512
Application Date: 20260209

Claims (12)

1. A method of identifying a modulated signal, the method comprising: acquiring a first signal of a modulation type to be identified; Inputting the first signal into a signal type identification model to obtain the modulation type of the first signal output by the signal type identification model; the signal type recognition model is used for outputting the modulation type of the signal according to the received signal and the signal; the signal type recognition model is obtained by pre-training an initial model for carrying out modulation type recognition on signals and carrying out migration training on the initial model.
2. The method according to claim 1, wherein the method further comprises: The universal software radio level platform USRP receiving end is utilized to acquire the second signal and the modulation type corresponding to the second signal; Inputting a third signal into the initial model, and extracting features of the third signal by using a convolution layer of the initial model to obtain a first signal feature, wherein the third signal is a signal with a first preset proportion in the second signal; carrying out pooling treatment on the first signal characteristics by utilizing a pooling layer of the initial model to obtain second signal characteristics; Processing the second signal characteristics by utilizing a long-short-term memory network LSTM layer of the initial model to obtain third signal characteristics of the last time step of the LSTM layer; obtaining a predicted modulation type according to the third signal characteristic by using an output layer of the initial model; and modulating model parameters in the convolution layer, the pooling layer and the LSTM layer according to the predicted modulation type and the modulation type corresponding to the third signal to obtain a pre-trained initial model.
3. The method of claim 2, wherein the convolution layer comprises a two-dimensional, two-dimensional convolution structure; the feature extraction of the third signal by using the convolution layer of the initial model, to obtain a first signal feature, includes: And carrying out joint extraction on the real part time-frequency characteristic and the imaginary part time-frequency characteristic of the third signal by utilizing the double-layer two-dimensional convolution structure to obtain the first signal characteristic.
4. A method according to claim 2 or 3, characterized in that the method further comprises: Verifying the pre-trained initial model by using a fourth signal and a modulation type corresponding to the fourth signal; wherein the fourth signal is a signal other than the third signal in the second signal.
5. The method according to claim 1, wherein the method further comprises: acquiring a fifth signal of different signal parameters under an analog channel environment and a modulation type corresponding to the fifth signal; performing migration training on a convolution layer and a pooling layer of the initial model by using training data to obtain the signal type identification model; The training data comprises a second signal with a second preset proportion, a modulation type corresponding to the second signal with the second preset proportion, a fifth signal with a third preset proportion and a modulation type corresponding to the fifth signal with the third preset proportion.
6. The method of claim 5, wherein the convolution layer comprises a two-dimensional, two-dimensional convolution structure; performing migration training on a convolution layer and a pooling layer of the initial model by using training data to obtain the signal type identification model, wherein the method comprises the following steps: Adjusting model parameters of a first convolution structure and model parameters of the pooling layer by utilizing the training data to obtain the signal type identification model; Wherein the first convolution structure is a last layer convolution structure in the double-layer two-dimensional convolution structure.
7. The method of claim 5, wherein the method further comprises: And replacing an output layer in the initial model with a full-connection layer, and training the full-connection layer by utilizing the training data to obtain the signal type identification model.
8. The method of claim 5, wherein the method further comprises: monitoring the signal type identification model by using verification data to obtain loss information; Determining whether the transfer learning of the signal type recognition model is stopped according to the loss information; The verification data comprises a fourth preset proportion of the second signal and a fourth preset proportion of a modulation type corresponding to the second signal.
9. A modulated signal recognition apparatus, the apparatus comprising: The first acquisition module is used for acquiring a first signal of a modulation type to be identified; The first processing module is used for inputting the first signal into a signal type identification model to obtain the modulation type of the first signal output by the signal type identification model; the signal type recognition model is used for outputting the modulation type of the signal according to the received signal and the signal; the signal type recognition model is obtained by pre-training an initial model for carrying out modulation type recognition on signals and carrying out migration training on the initial model.
10. A modulated signal recognition device comprising a processor, a memory and a program stored on the memory and executable on the processor, which when executed by the processor, implements the steps of the modulated signal recognition method according to any one of claims 1 to 8.
11. A readable storage medium, characterized in that the readable storage medium has stored thereon a program which, when executed by a processor, implements the steps in the modulated signal recognition method according to any one of claims 1 to 8.
12. A computer program product comprising computer instructions which, when executed by a processor, implement the steps in the modulated signal identification method of any of claims 1 to 8.

Description

Modulated signal recognition method, apparatus, device, medium, and program product Technical Field The present application belongs to the field of wireless technology, and in particular, relates to a method, apparatus, device, medium and program product for identifying a modulation signal. Background The modulation recognition technology has important application value. Modulation recognition techniques can be categorized into likelihood ratio-based modulation recognition and signal feature-based modulation recognition. The modulation recognition based on the signal characteristics is realized by extracting characteristic parameters or characteristic sequences (including frequency spectrum characteristics, power spectrum density, phase change, instantaneous amplitude, modulation index and the like) of the signal, and then judging the modulation type of the signal according to the characteristics. Because likelihood ratio-based methods require accurate establishment of likelihood ratio models of modulation types, challenges in model construction may be faced to complex modulation types, and signal feature-based methods are more flexible and can adapt to different modulation types according to different feature parameters, but require proper feature extraction and decision rule design, so that signal feature-based modulation recognition techniques are generally more widely applied. The traditional modulation recognition technology based on signal features firstly designs and extracts the signal features, and then designs a classification rule to classify the signal features in a modulation mode. Disclosure of Invention The embodiment of the application provides a modulation signal identification method, a device, equipment, a medium and a program product, which are used for solving the problem that the existing judgment of the modulation type of a signal by using manually designed signal characteristics and classification rules has low accuracy. In a first aspect, an embodiment of the present application provides a method for identifying a modulation signal, where the method includes: acquiring a first signal of a modulation type to be identified; Inputting the first signal into a signal type identification model to obtain the modulation type of the first signal output by the signal type identification model; the signal type recognition model is used for outputting the modulation type of the signal according to the received signal and the signal; the signal type recognition model is obtained by pre-training an initial model for carrying out modulation type recognition on signals and carrying out migration training on the initial model. Optionally, the method further comprises: The universal software radio level platform USRP receiving end is utilized to acquire the second signal and the modulation type corresponding to the second signal; Inputting a third signal into the initial model, and extracting features of the third signal by using a convolution layer of the initial model to obtain a first signal feature, wherein the third signal is a signal with a first preset proportion in the second signal; carrying out pooling treatment on the first signal characteristics by utilizing a pooling layer of the initial model to obtain second signal characteristics; Processing the second signal characteristics by utilizing a long-short-term memory network LSTM layer of the initial model to obtain third signal characteristics of the last time step of the LSTM layer; obtaining a predicted modulation type according to the third signal characteristic by using an output layer of the initial model; and modulating model parameters in the convolution layer, the pooling layer and the LSTM layer according to the predicted modulation type and the modulation type corresponding to the third signal to obtain a pre-trained initial model. Optionally, the convolution layer comprises a two-dimensional convolution structure of two layers; the feature extraction of the third signal by using the convolution layer of the initial model, to obtain a first signal feature, includes: And carrying out joint extraction on the real part time-frequency characteristic and the imaginary part time-frequency characteristic of the third signal by utilizing the double-layer two-dimensional convolution structure to obtain the first signal characteristic. Optionally, the method further comprises: Verifying the pre-trained initial model by using a fourth signal and a modulation type corresponding to the fourth signal; wherein the fourth signal is a signal other than the third signal in the second signal. Optionally, the method further comprises: acquiring a fifth signal of different signal parameters under an analog channel environment and a modulation type corresponding to the fifth signal; performing migration training on a convolution layer and a pooling layer of the initial model by using training data to obtain the signal type identification model; The t