This thesis establishes an automatic classification program for the signal detection work in pipeline inspection. Time-scale analysis provides the basic methodology of this thesis work. The wavelet transform is implemented in the program for filtering out the majority of noise and detect needed signals. As a popular nondestructive test, acoustic emission (AE) testing has been widely used in many physical and engineering fields such as leak detection and pipeline inspection. Among those applied AE tests, a common problem is to extract the physical features of the ideal events, so as to detect similar signals. In acoustic signal processing, those features can be represented as joint time frequency distribution. However, classical signal processing methods only give global information on either time or frequency domain, while local information is lots. Although the short-time Fourier transform (STFT) is developed to analyze time and frequency details simultaneously, it can only achieve limited precision. Other time-frequency methods are also applied in AE signal processing, but they all have the problem of resolution and time consuming. Wavelet transform is a time-scale technique with adaptable precision, which makes better feature extraction and detail detection. This thesis is an application of wavelet transform in AE signal detection where various noise exists. The wavelet transform with Morelet wavelet as the mother wavelet provides the basis of the program for auto classification in this thesis work. Finally the program is tested with two industrial projects to verify the workability of wavelet transforms and the reliability of the developed auto classifiers.