Abstract: As to the problem that the traditional covariation method has an infinite variance,which leads to a poor effect in practice in symmetric stable distribution noises(SαS), this paper proposes a covariation time delay estimation method based on signal preprocessing. The theoretical analysis indicates that the covariation of the new method has a finite variance by passing the received signals through any odd symmetry,monotone increasing and bounded functions,which improves the precision accuracy of time delay estimation by a large margin,making the apply of the method in practice become possible.At last two preprocessing methods which satisfying the conditions above are proposed.Taking the proposed preprocessing methods and the pre-existing arctangent method into accout,the simulation results prove that the proposed methods can improve the estimated accuracy in symmetric stable distribution noises(SαS) and can be used under Gaussian conditions.