Opportunity at Naval Air Warfare Center Weapons Division (NAWCWD)
Nonlinear Stable Signal Processing Using Invariant Scattering
Naval Air Warfare Center Weapons Division Research Dept., Physics Division
||China Lake, CA 93555
|Canning, Francis X.
Signal processing using Fourier Transforms generates a representation that is not stable with respect to properties of interest. For example, as longer time series are examined, Fourier coefficients become very dense in frequency. However, due to "time warping" (e.g., that resulting from a changing Doppler frequency) small physical changes (e.g., in Doppler) cause large changes in the Fourier coefficients. A nonlinear unitary transform has recently been developed (Refs 1 and 2) that provides coefficients that are stable with respect to parameters of physical interest. The opportunity is to explore using this method for numerous types of classification and identification problems. For example, can it be used to identify the types and numbers of aircraft present from their engine noises? Can aircraft be identified from a time series of the strength (magnitude) of their radar reflection? Can it be used to better receive transmitted information (voice or data) in the presence of noise?
Mallat S: Communications on Pure and Appl Mathematics 65(10):. 1331-1398, 2012
Anden J, Mallat S: Deep scattering spectrum, arXiv: 1304.6763 and submitted to IEEE Trans. On Signal Processing 2013./p>
Signal processing; Wavelet; Convolutional networks; Acoustics; Radar; Mel-frequency specrogram; Classification; Identification; Phase retrieval;
Open to U.S. citizens
Open to Postdoctoral applicants