Robust method for interference identification in new environment

By pre-training during the silent period and fine-tuning parameters during the non-silent period, the problem of identifying interference signals under communication signals is solved, achieving high-accuracy interference identification in the uplink of low-Earth orbit satellites, and adapting to changes in channel environment and noise.

CN115249025BActive Publication Date: 2026-06-05UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
UNIV OF ELECTRONICS SCI & TECH OF CHINA
Filing Date
2021-10-18
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing technologies are not ideal for identifying interference signals when there are communication signals (non-quiet periods), especially for low-Earth orbit satellite uplink interference signals with low accuracy, and the identification performance is very poor, especially in the absence of interference.

Method used

We employ Fine Tune (FT) and pre-adaptation methods from transfer learning. By pre-training the interference recognition model during the silent period and fine-tuning the parameters during the non-silent period, we use a small number of target domain samples for training to adjust the recognition model to adapt to the new environment.

Benefits of technology

High-accuracy identification of interference signals was achieved in the new environment, especially with a significant improvement in the identification accuracy in the absence of interference. The average identification accuracy reached 95%, and the identification accuracy of various types of interference was close to 100%, adapting to changes in channel environment and noise.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115249025B_ABST
    Figure CN115249025B_ABST
Patent Text Reader

Abstract

The application discloses a robustness method for interference recognition in a new environment and belongs to the field of signal interference. The application firstly sets an interference recognition model based on a convolutional neural network, adopts collected signals in a quiet period as a first training set, and performs basic training on the interference recognition model; then adopts collected signals in a non-quiet period as a second training set, performs second training on the interference recognition model trained in step 2, and takes the interference recognition model trained in the second time as an interference recognizer; wherein the loss function and the optimization function used in the second training and the basic training are the same, and the learning rate used in the second training process is lower than the learning rate used in the basic training; after pre-processing the signal to be recognized, the signal is input into the interference recognizer to obtain a recognition result. The application has good interference recognition effect, and is suitable for interference recognition in a new environment, such as when the channel environment changes and the noise changes.
Need to check novelty before this filing date? Find Prior Art