The biological signal registers (biosignals) are directly associated with states of the patient health, due to the dynamical variations of a biological system are usually related with changes in the observed state. Therefore, the accurate register and interpretation of those recordings, along with precise clinical information, are important tools to optimize its service relevance and utility for the control and diagnosis of pathologies, consequently a correct interpretation of the biosignals is directly related with the patient treatment. Nevertheless, the develop of algorithms for helping the support of clinical diagnosis is highly conditioned for the non-stationarity of the recordings, giving rise to several approaches to deal with this issue. In a local context, the Signal Processing and Recognition Group (DSP & CG) from Universidad Nacional de Colombia - Manizales, has focussed its labor in the automatic analysis of the biological activity, in order to propose methodologies to support the development of automatic systems for diagnosis assistance. Nowadays, the research group is interested in the analysis of brain activity with aim to detect patterns that may characterize among several kind of activities, to support clinical decision systems and rehabilitation procedures, as well. In this sense, huge efforts have been focussed in the research of functional neuroimages, which represent in a noninvasive way, the dynamics of the brain functions in normal and pathological states. Usually, the neural activity is estimated by the functional Magnetic Resonance Imaging - fMRI, which estimates directly the neuron dynamics with high spatial resolution but with low temporal resolution. Moreover, neural activity can be measured by means of electroencephalographic - EEG or magnetoencephalographic - MEG recordings (M/EEG), with a lower spatial resolution than fMRI, but with higher temporal resolution, being the neuroimaging approach more commonly used. Nevertheless, the correct analysis and interpretation of those recordings are conditioned by strong assumptions about the temporal behavior of the data, making its accurately representation anopen issue.