Our goal is to organize the ElectroEncephaloGram (EEG) signal so as to describe and image various brain activities. Our work is based on a data structure, a graph, which sums up the brain activity in the spatial, temporal and frequency domains. From the information contained in the time-frequency map of EEG signals, a graph is constructed. In order to analyze the complexity of the signal, our method is based on a multi-scale approach with several levels of information extraction. To compare different EEG signals, we use techniques of graph-matching with our data structure. The developed algorithm is based on the A* algorithm that allows us to compare variations of the recorded EEG in term o f latency, frequency, energy and activated areas. The results of this project show first, that the graph is an appropriate tool to reduce the cortical activity complexity, and second, that graph-matching offers some interesting perspectives in order to describe functional brain activity.