Traditional second order-based EP latency change detection methods usually degenerate under fractional lower order -stable noise conditions.This paper proposes a fractional lower order covariance based algorithm

which changes the noise contaminated EP signal into a finite second order moment process by a non-linear transform

and guarantees its convergence in the entire range of

α

∈(0

2

]

and obtains a high robustness and accuracy for latency change estimation.The paper also gives the theoretical analysis for the performance of AFLC and simulation and data analysis results.