This　paper　describes　a　frame　erasure　concealment　method　based　on　the　fractal　prediction,　which　is　proposed　based　on　the　fractal　structure　of　speech　signal;　it　can　mitigate　the　adverse　impact　of　frame　erasure　on　the　reconstruction　quality.　The　erased　frame　is　concealed　by　previous　multi-frame　information.　Therefore,　the　extra　delay　is　not　necessary　for　this　frame　erasure　concealment　method.　This　method　is　used　in　ITUT　G.722.1　coding　algorithm.　The　performance　evaluation　of　objective　PESQ　of　ITU-T　indicates　that　the　reconstruction　speech　quality　is　improved　while　the　frame　is　erased.