Project DetailsIn this work we studied the problem of identification and classification of flying objects using the Micro Doppler effect in RADAR signals, and suggested a new solution based on Manifold Learning. The Micro Doppler effect describes the shift in frequency of RADAR signals reflected from the different parts of the objects that are moving in relation to the center of mass of the target. We studied about the common solutions to this problem and implemented a Matlab simulation that mimics reflection of RADAR signals for testing purposes.
We then implemented the solution described in this work, where we perform a non-linear dimensionality reduction using Diffusion Map, interpolate the mapping for new points using Laplacian Pyramid Extension, and train an SVM classifier.
We compared this solution to the common solution using the simulated data and using data from RADA that was recorded on real RADAR systems. We investigated and implemented different ways to improve the basic algorithm and found that the new solution performs better on real data compared to the common solution.
