The Feasibility of Pulse Signal Classification by Spectral Parameters.
NAVAL POSTGRADUATE SCHOOL MONTEREY CALIF
Pagination or Media Count:
The paper explores the feasibility of fast transform coefficients as classification features for pulse type signals. The fast transforms investigated are Fourier FFT, Walsh FWT, and Haar FHT. A synthesized signal base containing 79 distinct pulse shapes of similar duration is analyzed for classification information compactness in the discrete time, Fourier, Walsh, and Haar bases. Nonparametric information measures are used. It is concluded that a Fourier basis representation enables the significant reduction of dimensionality necessary for further study as a generator of classification features. Author