In order to monitor the wear states of the tool during friction stir welding process online, we designed a set of acoustic emission online monitoring system based on the virtual instrument technology. The system, which combined with acoustic emission sensors, data acquisition card and signal conditioner, can collect acoustic emission signal during the friction stir welding process. In the experiment, 7075 aluminum alloy was welded by the needle stirring tool with screw thread, and the acoustic emission signal during welding was collected by the self-developed online monitoring system. Then the collected acoustic emission signals were processed by wavelet packet decomposition, the percentage of each frequency band after decomposition was calculated, and the energy distribution rule was extracted as the signal feature. The result shows that friction stir welding tools have different acoustic emission signal characteristics under different wear conditions. And the wavelet packet decomposition shows that when the wear of the mixing head is slight, the proportion of energy in the low frequency band is high; on the contrary, the energy of the low frequency band accounts for a relatively low proportion when the wear of the mixing head is serious.