The application of speech recognition techniques to radar target Doppler recognition: a case study

Pulse Doppler Radar Target Recognition using a Two-Stage SVM Procedure

“…In this paper, MFCCs are used as feature parameters to represent radar target echo signals as in [2] and [3]. In this section the real cepstrum is reviewed first.…”

“…The logarithmic spacing of bands also provides a more descriptive representation of PDR target echo signals because most of signal energy is in low frequency bands as shown in spectrogram plots (Figs. [1][2][3][4][5]. Computation of MFCCs is carried out using the discrete cosine transform (DCT) of a log power spectrum on a nonlinear mel-scale of frequency domain [9] as follows:…”

“…Therefore, an automatic classification system will be an important improvement and will provide valuable support for ground surveillance pulse-Doppler operators [2,3]. In [2] preliminary results of radar target recognition using speech recognition based methods [4] are reported.…”

“…In [2] preliminary results of radar target recognition using speech recognition based methods [4] are reported. In [3] Doppler signatures and cepstrum feature parameters are classified using hidden Markov models (HMMs) and Gaussian mixture models (GMMs) and an average recognition rate of 88% is achieved.…”

“…The surveillance space projections of these detections are input to the target tracking filter employed, and in this paper these projections are referred to as the (target tracking) measurements. Measurements may also have additional components or features [1][2][3], including the amplitude or the Doppler velocity. Proposed clutter measurement density estimator does not preclude measurement features from being used in the usual manner.…”