In this paper, the amplitude (ΔI d /I d ) distribution of random telegraph noise (RTN) induced by each trap in nanoscale devices is investigated based on the statistical experimental results. The RTN states are extracted through the proposed Gaussian mixture model (GMM). Mont-Carlo simulation is performed to extract the most probable results for mean trap number and each RTN amplitude. The results show that the RTN amplitude distribution is well consistent with the lognormal distribution instead of the exponential distribution for both the DC and AC results, which is helpful for future robust digital circuit design against RTN.