Classification of low probability of intercept radar waveforms using gabor wavelets Dinlenmesi düşük olasılıklı radar dalga biçimlerinin gabor dalgacık dönüşümü kullanarak sınıflandırılması
Journal of the Faculty of Engineering and Architecture of Gazi University, cilt.36, sa.4, ss.2025-2035, 2021 (SCI-Expanded, Scopus, TRDizin)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 36 Sayı: 4
- Basım Tarihi: 2021
- Doi Numarası: 10.17341/gazimmfd.782311
- Dergi Adı: Journal of the Faculty of Engineering and Architecture of Gazi University
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Art Source, Compendex, TR DİZİN (ULAKBİM)
- Sayfa Sayıları: ss.2025-2035
- Anahtar Kelimeler: Electronic support systems, Gabor wavelet transform, LPI radar, Waveform classification
- Ankara Yıldırım Beyazıt Üniversitesi Adresli: Evet
Özet
Purpose: In this study, the purpose is to classify LPI radar waveforms by comparing their features extracted using Gabor Wavelets. Unlike many methods given in the literature our method does not require learning phase and reshaping of input signals which cause information loss. Theory and Methods: To recognize 13 different LPI waveforms (12 different modulation types and non-modulation), signals are sampled at a sampling rate of 20 kHz. Then, CWD is applied to the sampled data. The output CWD-TFI is automatically cropped according to the high energy regions on the image. Then, a total of 40 filters with 5 different frequencies and 8 different orientations have been calculated. TFIs are filtered by these filters to extract feature vectors. The classification can be done by finding similarity between feature vectors of test signals and feature vectors of signals in database. Then the matching score is calculated to assign a label to tested signals. Results: For the classification of LPI waveforms, a total of 13 different signal types have been produced by taking SNR values between -20 dB and 10 dB in 2 dB intervals. In the figure below, the overall classification accuracy of the proposed method. Conclusion: One of the advantages of our method over the methods given in the literature is that there is no need to resizing or re-shaping of different LPI waveforms to make the images the same size. The emphasis here is not on just filling in zero or the time spent on it, but attempts to reduce the size of the images to avoid computational complexity results the loss of distinctive features, especially for LPI signals with low SNR values. Another advantage of our method is that there is no learning phase, so classification can be done using feature vectors saved in the database.