This is a proposal to develop new and improved techniques that might be of special use in analyzing time-varying biological signals. The approach is to design appropriate kernels to be used in time-frequency representation of signals, a representation that is a generalization of the classical Fourier transform. The new method is expected to provide an improved description of signals whose characteristics vary with time. The proposed research will investigate the use of these techniques for the analysis of biomedical signals: EEG in epileptic patients, joint sounds in patients with joint problems, and evoked potentials in phobic patients. The research, if successful, will result in a set of new analytic techniques for characterizing signals, and the demonstration of the usefulness of these techniques for diagnostic purposes.