《Digital-signal-processing technique for ultrasonic signal modeling and classification》.pdfVIP

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 50, NO. 1, FEBRUARY 2001 15 A Digital-Signal-Processing Technique for Ultrasonic Signal Modeling and Classification Angelo M. Sabatini, Member, IEEE Abstract—In this paper, we face the problem of constructing Because of these limitations, in-air acoustic imaging systems a robust feature space for automatic classification of signals from have still quite limited applications for the purpose of pattern narrow-band in-air ultrasonic sensors. In consideration of the ex- recognition and scene understanding. Current systems work on isting sensor bandwidth restrictions, the importance of selecting a suitable signal descriptor is highlighted. We assume that the char- range sonar images which are generated using either a single acteristics of the ultrasonic sources which produce the signals are sensor, mechanically swept in a sequence of rasters, or a sensor impressed in the shape of their echo envelopes. array whose beam is electronically deflected [3]–[5]. Mostly, A technique based on orthonormal Laguerre polynomials is ap- the identification process follows the reconstruction of a suit- plied to the echo envelopes for constructing the feature space. Dif- able spatial image of the object. However, object identification ferent methods for computing the Laguerre coefficients are dis- cussed, and the properties of the resulting feature space are inves- can also be based on the information extracted from a single tigated. receiving sonar sensor at a fixed position [6]–[10]. The anal- For the experimental verification of the metho