Dynamic wavelet fingerprint identification of ultrasound signals

This paper presents an automatic ultrasound signal classification technique based on a dynamic wavelet fingerprint. Dynamic wavelet fingerprints are constructed on the spot from a projection of the continuous wavelet coefficients of the transient signal, which converts a one dimensional signal classification problem into a two dimensional pattern recognition problem. Features are then extracted in terms of the moment invariants of the dynamic wavelet fingerprint, which are used as inputs to an artificial neural network to identify different transient signals. The identification performance is tested with an application to ultrasonic sorting materials.