Abstract:Radiation source signal sorting in complex environments is currently a hot issue in the field of electronic countermeasures. The pulse repetition interval (PRI) can provide stable and obvious parametric features in radiation source identification, which is an important parameter relying on the signal sorting problem. To solve the problem linked to the difficulties in sorting the PRI in complex environments using the traditional method, a signal sorting method based on a parallel denoising autoencoder is proposed. This method implements the binarized preprocessing of known time-of-arrival (TOA) sequences and then constructs multiple parallel denoising autoencoder models using fully connected layers to achieve the simultaneous sorting of multiple signal types in the overlapping signals. The experimental results show that this method maintains high precision in scenarios prone to large error and can efficiently filter out noise and highlight the original features of the signal. In addition, the present model maintains its performance and some robustness in the sorting of different signal types. Compared with the traditional algorithm, this method improves the precision of sorting. The algorithm presented in this study still maintains above 90% precision when the pulse loss rate reaches 50%.
engineering, electrical & electronic,computer science, information systems,physics, applied
What problem does this paper attempt to address?
This paper proposes a new solution to the problem of identifying and classifying radiation source signals in complex environments. Specifically, the research focuses on how to effectively classify pulse repetition intervals (PRI) in the presence of significant errors, missing pulses, and false pulses.
### Research Background and Problem Definition
- **Background**: With the widespread application of radar signals in various fields, signal interference and parameter space overlap in complex electromagnetic environments are increasing, posing challenges to the identification of radar radiation sources.
- **Problems with Traditional Methods**: Traditional five-dimensional parameter methods (based on parameters such as carrier frequency, pulse amplitude, pulse width, direction of arrival, and time of arrival) are prone to parameter ambiguity in complex environments, affecting the accuracy of signal classification.
- **Key Parameter**: The paper points out that the pulse repetition interval (PRI) is a stable parameter with distinct pattern characteristics, making it an important basis for signal classification.
### Solution Overview
- **Main Contributions**:
- A method based on parallel denoising autoencoders is proposed, which can achieve multi-target parallel classification tasks.
- Binary encoding is used to preprocess the original time of arrival (TOA) sequence to improve data adaptability and model recognition efficiency.
- A network model containing multiple parallel-working denoising autoencoders is constructed, suitable for signal identification and classification in noisy environments.
### Methodology
- **Data Preprocessing**: Binary encoding is used to preprocess the TOA sequence, converting the arrival time of each pulse into a binary sequence for subsequent processing.
- **Model Architecture**: A model is constructed using denoising autoencoders (DAE), which can add artificial noise to the input data to train more robust denoising capabilities.
- **Parallel Model**: A parallel model containing multiple independently working denoising autoencoders is designed to handle multiple signal types simultaneously, improving the efficiency and accuracy of signal classification.
### Experimental Design
- **Sample Simulation**: Four typical modulation types (fixed repetition frequency, sliding repetition frequency, pulse group repetition frequency, and periodic modulation repetition frequency) were selected as experimental samples.
- **Error Simulation**: Three types of errors—measurement errors, missing pulses, and false pulses—were introduced to simulate signal interference in actual environments.
- **Network Parameter Settings**: Based on the experimental samples, a network model containing four autoencoder units was established, with each unit having detailed parameter configurations.
### Conclusion
The proposed method, by combining preprocessing techniques and a parallel denoising autoencoder model, demonstrates higher accuracy and robustness in handling signal classification tasks in complex environments compared to traditional algorithms. Notably, the method maintains over 90% classification accuracy even in the face of high missing pulse rates, highlighting its significant application value.