What have I done on my PhD?

Mais informações

Macroeconomics - 2HH3 - Chapter 3

por Remek Debski

The numbers

por Gabriela Casado Patiño

Machine Learning Algorithms

por 蔡 蔡

The boy who tamed the wind

por Joffre Veloz

What have I done on my PhD?

Started to consider what is meant by multiperspectiveclassification in the case of the uDS.

Investigated how clutter can be removed from databeing processed

Used a method based on the CLEAN algorithms

Used the zero Doppler removal technique fromturntable ISAR.

Gathered a set of multistatic data from trials at Shenley

Work out how to process the data

Fusing of the data from all nine bistatic paths

This is still in progress

Cration of the reconstructed signals

The range-FFT plots

Conducted the trials

Measured the biometric data of persontargets.

Gathered the data

Map of the range

Planned out the trials

Decison of what target to investigate

Decision on radar parameters to use

Concerns over multipath

Consideration of SNR/SCR

Classification of Thales data usingother techniques

Artificial neural network

Hidden Markov Model

Bayesian classification

Statistical classification/kernal estimation

Frequency domain processing of data

This is currently in progress for Thales data

Analysis of classification output

Contrasted the impact of adding extra perspectivesto the library for the classifier.

The crazy plots of parameter variationwith frame-duration and correlation-level

The performance evaluation parameters

R

Pgen

Pfa

Pd

Pcc

Used ROC curves to assess the unknown thresholds

Time domain processing of data

Classification of Thales data using k-nn

Explored two types of library

Monperspective

Implemented a version of the k-nn algorithm

Classification of Thales data using DTW

Explored three types of library

Sparsely populated

Biperspective

Monoperspective

Implemented two version of DTW

Need to implement can't declare

Allowed for unknown input

Type II

Type 0

Investigated how the library and testdata can beselected based on correlation distance of thesignal.

The results using these libraries should becompared with the conventional randomselection method that is often used.

More investigation of this is required baseon the idea that k-nn forced me to add extraentries to the library.

Developed a set of libraries based on thedecorrelation distance.

Assessed the correlation distance of theseframes.

Broke the data into a number of sub framesto better emulate a radar's dwell time.

Analysis of the Thales data

Feature extraction from the data

Looked at how this may be analysed through FFT & STFT

Processed the data to form the reconstructed target signal

Developed a procedure for reading the data into matlab

Produced a basic simulation of the micro-Doppler effect

Matlab simulation of walking person

Matlab simulation of propeller

Matlab simiulation of multiple vibrating scatterers