A method for dynamic selection of channel in a wireless sensor network is presented in this paper. It is a simple cognitive mechanism to learn the behavior of operational band and to access channels with lower interference levels. The method is evaluated in term of the learning agility. The sink node collects energy detection measurements from sensor nodes to feed a reinforcement learning, which measures the interference energy over the channels. The results are analyzed in terms of learning evolution, operational channel evolution, primary symbols interfered, data and bit errors in the secondary nodes. The method achieves to learn the primary behaviors on each channel simply and fast. Benefits in terms of spectral efficiency and coexistence are reached using the proposed method.