An Algorithm for Time Prediction Signal Interference Detection Based on the LSTM-SVM Model.

Area of Science:

• Signal Processing
• Machine Learning
• Electronic Defense Systems

Background:

• Traditional interference detection methods struggle with signals overlapping in time and frequency.
• Detecting interference within the same frequency as the original signal presents significant challenges.

Purpose of the Study:

• To propose a novel interference detection algorithm for time-frequency overlapping signals.
• To enhance the accuracy and capability of electronic defense systems in identifying complex interference patterns.

Main Methods:

• Utilized a long short-term memory (LSTM) network for time series prediction of received signals.
• Employed support vector machines (SVM) for classifying feature samples derived from signal prediction differences.
• Compared the proposed LSTM-SVM model against a gate recurrent unit-support vector machines (GRU-SVM) model.

Main Results:

• The LSTM-SVM model successfully detected the presence of interference signals.
• The algorithm accurately determined the specific position of interference within the received waveform.
• Performance evaluation demonstrated superior detection capabilities compared to the GRU-SVM model, visualized via confusion matrices.

Conclusions:

• The proposed LSTM-SVM algorithm offers a robust solution for detecting and locating challenging time-frequency overlapping interference.
• This approach significantly improves upon existing methods, enhancing electronic defense system effectiveness.
• The model's ability to pinpoint interference location provides critical information for signal management.