Non-cooperative target recognition (NCTR) means technologies that provide high-resolution target signatures and use them to make some decisions about the type of target that has been detected. Clearly, this function is performed with no cooperation from the target concerned and it may even not be aware that its signature is being measured for recognition purposes. Jet engine modulation (JEM) has been widely used as a representative NCTR method by providing unique information on targets. JEM, induced by electromagnetic scattering from a rotating jet engine compressor, is one of the micro-Doppler phenomena that impart fre-quency modulation to radar signals. In this dissertation, automatic feature extraction of JEM signals is intro-duced in radar signal processing in order to improve real-time NCTR performance of JEM.
This dissertation is composed of four main parts: Improved algorithm for estimating the fundamental periodicity of JEM, advanced joint time-frequency analysis (JTFA) based on image processing method, fea-ture extraction from insufficient JEM signals based on compressed sensing (CS) method and novel feature extraction for measured insufficient JEM signals based on modified empirical mode decomposition (EMD).
In the first part, an improved algorithm is presented for automatically extracting the rotation period of the jet engine, also called as the spool rate. First, wavelet decomposition (WD) with Meyer wavelet is applied to the analytic form of the JEM signal. Then, decomposition components are combined to have a sufficient energy and minimize the mean of the JEM auto-correlation. Finally, the peak detection algorithm is em-ployed for automatic estimation of the spool rate. Application results of the simulated and measured JEM signals demonstrated that the proposed automatic algorithm is robust to noise, effective in accurate and fast estimation of the spool rate and has good applicability for a variety of JEM signals.
In the second part, an advanced JTFA for feature extraction is presented based on image processing method. First, EMD with adaptive low-pass filtering is employed to effectively extract the first harmonic component of the JEM signal. Then, the extracted component is used for reconstructing the JEM signal clari-fied in the joint time-frequency (JTF) domain. After converting the JTF representation into an image with RGB colours, the green component was extracted as a representative of the JEM component. Finally, the peaks detected from the extracted green component can represent the jet engine features. The new approach of JEM analysis is significant because the overall procedures for extracting the jet engine features are not manual but automatically performed based on the image processing method. The validity of the proposed method is demonstrated using measured JEM signals with complicated frequency composition
In the third part, an effective method for extracting JEM features is presented by reconstructing insuf-ficient JEM signal based on CS method. First, CS method is employed to extract a refined spectrum from insufficient JEM signals. In the process of applying the CS method to JEM signals, we utilized the modified linear programming (LP) standard form of an optimization method. After converting the spectrum into the time domain signal using the inverse Fourier transform, the cepstrum function was used for acquiring the spool rate. Then, the final blade number is estimated by introducing the divisor-multiplier (DM) rule and the scoring concept into JEM spectral analysis. Its application to various JEM signals showed that the proposed method enable reliable estimation of JEM features in spite of the insufficient JEM signal and is and is ex-pected to be effective from the perspective of radar resource management.
In the fourth part, a novel automatic algorithm is presented to estimate the blade number of meas-ured insufficient JEM signals. First, a modified EMD is employed to extract the first harmonic component of the JEM signal. Then, the decomposed intrinsic mode functions (IMFs) are combined to acquire the refined auto-correlation function (ACF). Finally, the blade number of a jet engine is estimated using the peaks de-tected from ACF. The proposed algorithm is innovative due to only using the time-domain method, not the frequency-domain method. The application of the proposed algorithm to measured insufficient JEM signals demonstrates that the novel automatic algorithm improves the accuracy of JEM analysis, and its application is expected to enhance the recognition performance with short measurement time.