To　reduce　the　noise　in　the　clock　frequency,　since　the　hydrogen　atomic　clock　frequency　differences　are　non-linear　and　non-stationary　with　time,　the　denoise　algorithm　based　on　ensemble　empirical　mode　decomposition　is　put　forward.　Firstly,　the　clock　frequency　differences　are　added　a　certain　intensity　noise;　then　decompose　the　differences　with　the　EMD　algorithm,　repeat　this　step　many　times;　finally,　calculate　the　mean　of　the　same　components　to　get　the　denoise　clock　frequency　differences.　Compared　with　the　wavelet　algorithm　in　the　time　domain　and　frequency　domain,　the　result　shows　that,　the　algorithm　put　forward　in　this　paper　is　better　than　the　wavelet　algorithm　in　the　denoise　of　the　clock　frequency　differences,　the　data　variance　is　0.7263%　which　is　2.707%　in　the　wavelet　algorithm,　the　data　become　more　stable.　©　2017,　Acta　Metrologica　Sinica　Press.　All　right　reserved.