The jet gas from the internal leakage in a ball valve, which is an indispensable equipment in a high-pressure natural gas transmission pipeline, can produce acoustic emission (AE) signals. Thus, the research on the characteristic rule of the AE signals will be helpful for the quantitative detection of the valve leakage flow. In dealing with this problem, the AE mechanism and testing of internal leakage in a ball valve are investigated. An AE testing system is used to measure the leakage in ball valves of three different sizes. In the analysis process of the testing data, the spectrum distribution characteristics of the AE signals on leakage levels are described, and the characteristic parameters of signals (information entropy, root mean square (RMS) and frequency peak) are extracted using the wavelet packet analysis method. On this basis, fitted curves about the characteristic parameters and the leakage flow are established, and the correlation is evaluated using the R-square (determination coefficient). The evaluation results show that the RMS value with the highest correlation (R-square =0.979) can be used for the quantitative detection of leakage in a ball valve in the natural gas pipeline.