Title :
A robust discriminant parameter set for the underwater object classification
Author :
Radoi, Emanuel ; Quinquis, André
Author_Institution :
EIA Dept., ENSIETA, Brest, France
Abstract :
The time-frequency analysis of the magnetic signals, generated by ferromagnetic objects, is used in order to find a robust discriminant parameter set for their classification. An original combination of energetic, spectral and time-scale based parameters is proposed. An excellent recognition rate and a good and regular behavior in the presence of noise are obtained. It is also proven that only a complex analysis of the acquired signals using their representations in different domains is able to lead to good results
Keywords :
feature extraction; neural nets; object detection; object recognition; pattern classification; signal sampling; singular value decomposition; spectral analysis; time-frequency analysis; vector quantisation; wavelet transforms; MUSIC algorithm; Nyquist sampling; SVD; complex analysis; discriminant analysis eigenvalues; energetic parameters; feature selection; ferromagnetic objects; learning VQ classifier; magnetic signals; neural net; parameter extraction; passive sensor; recognition rate; robust discriminant parameter set; spectral parameters; time-frequency analysis; time-scale based parameters; underwater object classification; wavelet packet analysis; Acoustic sensors; Magnetic analysis; Magnetic field measurement; Magnetic sensors; Noise robustness; Signal analysis; Signal generators; Signal processing; Signal to noise ratio; Time frequency analysis;
Conference_Titel :
OCEANS '97. MTS/IEEE Conference Proceedings
Conference_Location :
Halifax, NS
Print_ISBN :
0-7803-4108-2
DOI :
10.1109/OCEANS.1997.624093
