Self-organized criticality is being considered as a potential organization of the brain. In this study, major features of critical systems were applied to investigate the power-law distributions of human electrocorticogram (ECoG) data, with the aim of determining whether the critical regime could be applied to reveal the underling change of epileptic seizure generation. Multiple brain region ECoG signal was recorded from three epilepsy patients, including inter-ictal, pre-ictal, ictal and postictal stages. The Hurst exponent (H) parameter from the power-law analysis was calculated based on the ECoG signal spectrum estimated using a harmonic wavelet transform-based power-law analysis method. The changes in H at normal, inter-ictal spike, ictal stages were discussed in the framework of criticality theory. The H parameter could describe the dynamics of seizure generation. When inter-ictal spike occurred, H became larger than 0.5, suggesting that the underlying system changed from non-persistent to persistent dynamics. However, when seizure occurred, the ECoG dynamics changed into a state that H cannot indicate. The power-law analysis with Hurst exponent can be used to describe the generation of epileptic seizure. This analytical method provides a new insight to the understanding of the generation mechanism of epileptic seizures in terms of criticality, which could be used to design a prediction and/or detection method for closed-loop control of epilepsy.