In view of the low efficiency and poor coding performance, an active mechanism network coding algorithm based on multi-channel sensing is proposed. The algorithm first establishes a system model to analyze the work pattern of network coding, then selects the minimum sample as cost function and adopts the multi-channel sensing to examine the coding effect of relay node, which makes the network to learn from the coding effect of relay node so as to send the source information to the relay node with better encoding effect. In order to reduce encoding error rate, active mechanism is adopted to improve the capability of network decoding, where the mechanism uses the transmit power of source node and the size of source information to control terminal decoding error rate. The experiment and simulation results show that our proposed network coding algorithm has a very good effect on encoding and decoding processes and can effectively reduce the network overhead.