In this chapter, we propose a novel method for tracking oscillatory components in EEG signals by means of an adaptive filter-bank. The specificity of our tracking algorithm is to maximize the oscillatory behavior of its output rather than its spectral power, which shows interesting properties for the observation of neuronal oscillations. Besides, the structure of the filter-bank allows for efficiently tracking multiple frequency components perturbed by noise, therefore providing a good framework for EEG spectral analysis. Moreover, our algorithm can be generalized to multivariate data analysis, allowing the simultaneous investigation of several EEG sensors. Thus, a more precise extraction of spectral information can be obtained from the EEG signal under study. After a short introduction, we present our algorithm as well as synthetic examples illustrating its potential. Then, the performance of the method on real EEG signals is presented for the tracking of both single oscillatory component and multiple components. Finally, future lines of improvement as well as areas of applications are discussed.