A brief description is given of both a wideband echo-sounder and data acquisition at sea. Experiments were conducted in the Bay of Biscay at various seasons for four years. Fish species were identified by trawling. Only echoes associated with trawl catches that were monospecific were used in the classification analyses. Species discrimination was based only on the spectral signature of the echoes and did not take into account the characteristics of the school shape. A modelling of the power spectrum of the echo was used to limit the spectral signature to a reduced set of parameters that could be used for classification using a neural network. Thirty-six diVerent monospecific schools, including about 900 echoes, were processed. Three species were considered: sardine (Sardina pilchardus), anchovy (Engraulis encrasicolus), and horse mackerel (Trachurus trachurus). Classification performance had a success rate as high as 75%. ? 1996 International Council for the Exploration of the Sea
[1]
Harry L. Van Trees,et al.
Detection, Estimation, and Modulation Theory: Radar-Sonar Signal Processing and Gaussian Signals in Noise
,
1992
.
[2]
Steven Kay,et al.
Modern Spectral Estimation: Theory and Application
,
1988
.
[3]
A. Gray,et al.
Distance measures for speech processing
,
1976
.
[4]
R. Altes.
Sonar for generalized target description and its similarity to animal echolocation systems.
,
1976,
The Journal of the Acoustical Society of America.
[5]
Philip D. Wasserman,et al.
Neural computing - theory and practice
,
1989
.