Abstract The acoustic emission (AE) signals during four-point bending fatigue crack propagation in the base metal and weld of 16Mn steel were monitored and analyzed. The fatigue properties and acoustic emission characteristics were analyzed based on the micro-structural and fractographic observations. The results show that: AE signal was divided into three stages, correspond with the three stages of fatigue damage process, which were crack initiation, stable crack extension and unstable crack extension. The corresponding dominated AE source were micro-crack initiation in the first stage, stacking and slipping of dislocations ahead of the crack tip in the intermediate stage, shearing of ligaments and connectivity between dimples in the final stage. Fatigue crack growth rate and AE count rate is greater in weld specimen than in the base material, mainly due to the coarse ferritic grain in widmanstatten structure and a large number of oxide inclusions.The fatigue crack length was predicted by BP neural network method and AE parameter fitting method respectively. The relative error of each specimen was less than 15%. The results showed that acoustic emission technology could be used for fatigue life prediction and safety assessment in 16Mn steel and welds.