Empirical validation of the performance of a class of transient detector

Transient detection in the presence of noise is a problem which occurs in many areas of engineering. A description is given of a classifier system suitable for the identification of high frequency waveforms. It uses the wavelet transform for signal pre-processing and a forward feature selection algorithm. A radial basis function neural network is employed to model the class conditional probability density function. A short review of statistical pattern recognition is presented. The classifier is applied to the identification of a number of difficult to classify, high frequency acoustic emission signals