Application of signal processing and neural net techniques to ultrasonic characterization of multilayered media

The characterization of multilayered media using ultrasonic waves is classically performed using time of flight reflectometry in the pulsed regime or spectroscopy in the quasi-continuous wave regime. Our aim is to study the performances improvements given by numeric signal processing and neural nets methods. Starting from simple models for layered media, the reflection coefficient for high frequency pulsed ultrasonic waves is computed. The application of homomorphic processing techniques is studied theoretically, taking the transducer frequency response and the overall measurement system signal to noise ratio into account. Neural nets are then considered and it is shown that they provide a valid response in situations where other methods fail