In this paper, a new enrollment method for human identification using ECG is proposed. In the method, ECG data of a user are enrolled from five different poses while the user is holding a pair of dry electrodes from an ECG sensor with both hands. The five poses are when the user is holding the sensor in the center, left side, right side, upside, and downside of the user's body. The ECG data are collected from nine subjects and classification is performed by three existing algorithms and an original algorithm. Experiment results show that identification accuracy when ECG data are enrolled by the proposed method is improved from 3 to 13 percentage points than when they are enrolled in conventional single holding pose.