Method and device for waveform identification

Abstract

Method and device for waveform identification This method performs waveform identification from a digitized signal represented as an initial vector of T representative samples of the received electromagnetic waves.The process comprises the following steps: application (40) of a wavelet scattering network to the initial vector, and obtaining at output a number P of intermediate vectors (Vi_1,…,Vi_P), calculation (44) of a statistical vector for each intermediate vector, each statistical vector comprising predetermined statistical parameter values ​​calculated for said intermediate vector, formation (46) of a description vector (DV) of the initial vector from said statistical vectors, application (46) of a classification model taking as input the description vector, the classification model being previously trained by machine learning, to provide a prediction of membership in a waveform class among a set of N predetermined waveform classes. Figure for the abstract: Fig. 3.

Images