In order to improve NDE reliability, it is important to recover as much as possible of the useful information in NDE waveforms. An on-line minicomputer is ideally suited to both the collection of data and the performance of sophisticated signal processing tasks. Using a variety of signal processing techniques, including windowing, self-normalization (of transducer properties and far-field diffraction effects), transformations (Fourier magnitude and phase transforms, autocorrelations, cepstra), feature extraction and pattern· recognition, it has been possible to obtain information about very small defects, strength of adhesive bonds and acoustic emissions which are not available by conventional means. Examples of these various capabilities are given.