This work shows an approach to estimate brain activity using ensemble Kalman filter. The estimator is based on no linear model that considers the evolution of the neural activity as non-homogeneous. These conditions can help to improve the solution of the neuromagnetic inverse problem and to find spatially focus sources. This model allowed us to simulate brain activity similar to epilepsy in a specific area and differents time intervals, which could be observed in the multichannel electroencephalography, and is therefore a first step to carry on a focused analysis of the brain activity and the inter-connectivity among zones.