Pedestrian navigation systems based on smartphone are experiencing fast progress in indoor environment. Pedestrian dead reckoning approaches combined with improved inertial sensors' quality and the exploitation of magnetic field are used to mitigate the sensor drifts. The last remaining issue is related to the hand dynamic. It consists in estimating the angular misalignment between the smartphone pointing direction and the walking direction. Even though, some methods exist, their performances are lacking accuracy and reliability. A comparison of the three main methods to estimate this angular misalignment is performed. These methods are based on Principal Component Analysis (PCA), Forward and Lateral Accelerations Modeling (FLAM) and Frequency analysis of Inertial Signals (FIS). Despite better results for the FIS method all algorithm suffer from large outliers and a need for improved robustness is identified