What have I done on my PhD?
Produced a basic simulation of the micro-Doppler effect
Matlab simiulation of multiple vibrating scatterers
Matlab simulation of propeller
Matlab simulation of walking person
Analysis of the Thales data
Developed a procedure for reading the data into matlab
Processed the data to form the reconstructed target signal
Looked at how this may be analysed through FFT & STFT
Feature extraction from the data
Investigated how the library and testdata can beselected based on correlation distance of thesignal.
Broke the data into a number of sub framesto better emulate a radar's dwell time.
Assessed the correlation distance of theseframes.
Developed a set of libraries based on thedecorrelation distance.
More investigation of this is required baseon the idea that k-nn forced me to add extraentries to the library.
The results using these libraries should becompared with the conventional randomselection method that is often used.
Time domain processing of data
Classification of Thales data using DTW
Implemented two version of DTW
Type 0
Type II
Allowed for unknown input
Need to implement can't declare
Explored three types of library
Monoperspective
Biperspective
Sparsely populated
Need to implement can't declare
Classification of Thales data using k-nn
Implemented a version of the k-nn algorithm
Allowed for unknown input
Explored two types of library
Monperspective
Biperspective
Analysis of classification output
Used ROC curves to assess the unknown thresholds
The performance evaluation parameters
Pcc
Pd
Pfa
Pgen
R
The crazy plots of parameter variationwith frame-duration and correlation-level
Contrasted the impact of adding extra perspectivesto the library for the classifier.
Frequency domain processing of data
This is currently in progress for Thales data
Classification of Thales data usingother techniques
Statistical classification/kernal estimation
Bayesian classification
Hidden Markov Model
Artificial neural network
Gathered a set of multistatic data from trials at Shenley
Planned out the trials
Consideration of SNR/SCR
Concerns over multipath
Decision on radar parameters to use
Decison of what target to investigate
Conducted the trials
Map of the range
Gathered the data
Measured the biometric data of persontargets.
Work out how to process the data
The range-FFT plots
Cration of the reconstructed signals
Fusing of the data from all nine bistatic paths
This is still in progress
Investigated how clutter can be removed from databeing processed
Used the zero Doppler removal technique fromturntable ISAR.
Used a method based on the CLEAN algorithms