In this talk we will discuss the problem of testing the equality of two distributions by integrating the square norm of the difference between two corresponding empirical characteristic functions over a symmetric interval. This results in a linear combination of three different $U$-statistics. So the original testing problem is reduced to testing whether this linear combination is zero or not. Naturally we apply the jackknife empirical likelihood (JEL) method to the new hypothesis testing problem. It has been proved that, under multivariate case, the log JEL statistic after scaling tends to a chi square distribution with degree of freedom 1. Simulation studies are presented to study finite-sample performance of our method. This work is joint with Liu Yiming and Liu Zhi.