Title :
Using fractal interpolation function encoded ultrasonic signals to train a neural network
Author :
Ali, Maaruf ; Meng, Xiaoping ; Clarkson, Trevor G. ; Taylor, John G.
Author_Institution :
Dept. of Electron. & Electr. Eng., King´´s Coll., London, UK
Abstract :
This abstract describes novel research in using fractally compressed ultrasonic transducer signals using Fractal Interpolation Function (FIF) codes to train probabilistic RAM neural (pRAM) networks in order to differentiate between flawed and unflawed concrete blocks. Two classes of ultrasonic spectrum are used, one for the flawed concrete block and the other for the unflawed concrete block. The spectrum thus obtained is fractally encoded using the one dimensional form of fractal compression known as fractal interpolation functions
Keywords :
concrete; fractals; interpolation; learning (artificial intelligence); neural nets; ultrasonic imaging; fractal compression; fractal interpolation function; neural network; probabilistic RAM; train; ultrasonic signals; ultrasonic transducer signals;
Conference_Titel :
Fractals in Signal and Image Processing, IEE Colloquium on
Conference_Location :
London
DOI :
10.1049/ic:19950020
