A critical parameter of any automated coal mining system is the ability to cut only coal and not the rock and/or waste surrounding it. To do this, the system must be able to detect where the coal- rock-waste interface is. The U.S. Bureau of Mines is investigating a new approach to vibrational coal interface detection by utilizing an adaptive signal discrimination (asd) system to perform feature extraction and classification of complex vibrational signals. Accelerometers, geophones, and piezoelectric films are mounted on mining machines and/or strata and are used to monitor vibrations produced during mining operations. Selected data are digitized, and the asd system is trained on known types of vibrations (e.g., "cutting coal" and "starting to cut roof"). A signal classifier is then generated that can be used to identify unknown waveforms. The asd system can be retrained if mining conditions change significantly.