Pedestrian dead reckoning (PDR) is one of the most employed strategies to process inertial signals collected with a handheld device for autonomous indoor positioning. This strategy is based on step length models that usually combine step characteristics with some physiological parameters. These models are calibrated with experimental data for each user. However, many physiological conditions are affecting the walking gait even for steady walking. Therefore, frequent calibration is needed to cope with walking pattern variations. Moreover, PDR models are not adapted to high walking velocities and to the specific walking patterns of some populations like elderly people and pathological cases. In light of these limitations, the modeling of human walking, which considers the induced arm swinging behavior, is needed for improving self-contained inertial indoor navigation. In this paper, a human-like walking model is developed in order to represent and study the correlations between the hand acceleration and gait characteristics. Experimental data were collected from motion capture experiments on one healthy subject in order to validate the model. Results show that the model fitted to the test subject reproduces the walking features found in experiments, as well as the same tendencies in function of the walking velocity.