Вы находитесь на странице: 1из 1

Feature extraction and target Classification by micro-Doppler signatures with the Empirical Mode Decomposition (EMD)

CDS Visiting Research Student Programme 2014

Supervisor: Dr Alessio Balleri Abstract The echo produced by a moving target that is illuminated by a radar system contains frequency modulations caused by the time-varying delay occurring between the target and the sensor. The main bulk translation of the target towards or away from the sensor induces a frequency shift of the echo as a result of the well-known Doppler Effect. Additional movements of small parts of the target contribute with frequency modulations around the main Doppler shift and these are commonly called micro-Doppler modulations. MicroDoppler modulations contain a signature of the target that can be used for target recognition. Most of the work on target classification by micro-Doppler signatures is based on the use of the Short Time Fourier Transform (STFT) although it is well known this can only offer a good time resolution at the expenses of the frequency resolution (and vice versa). The Empirical Mode Decomposition is a new method to decompose non-stationary signals that can be used to study micro-Doppler signatures. This can potentially result in improved classification performance as it does not suffer from time-frequency resolution limitations. The objectives of this project are 1) to run a set of measurements with a real radar module to collect micro-Doppler signatures of various targets at a range of operating frequencies spanning from Sband to X-band, 2) to develop the Matlab software required to decompose micro-Doppler signatures using the EMD, and 3) to investigate whether classification performance of typical classifiers improves when EMD is used with respect to existing methods using the STFT. Very little work has been done in this area and this is an excellent opportunity to work on a real and timely problem with potential high research impact.
Keywords: Doppler Effect, radar systems, target classification, feature extraction.

Pre-requisites: A solid background on the theory of Fourier transforms and linear filters. Proposed Questions 1) Describe the definition and the meaning of Signal to Noise Ratio (SNR), what impact SNR may have on a receiver performance and what solutions can be adopted to maximise it. 2) Describe how observation time may affect the measured frequency components of an electromagnetic waveform. Comment on advantages and disadvantages of long observation times by providing real life examples.