Fluctuation of upright posture during quiet human standing, called postural sway, includes important information about the underlying postural control system. That is, size and temporal pattern of the sway change in a complex manner depending on individuals, age, and severity of movement disorders. One way to understand mechanisms of such changes is to consider a dynamic system model of postural control, in which mechanical dynamics of the standing body are stabilized by neuro-muscular controllers. The intermittent feedback control model, proposed by the authors, is one of such promising models. In this study, we assimilate the intermittent control model to human postural sway data based on the approximate Bayesian computation that utilizes a set of summary measures. Those measures are used to compute a distance between given two time-series data, the one acquired experimentally and the other simulated numerically by the model, whereby parameter values (a joint distribution of the parameter values)of the model are inferred so that the model could generate a simulated time series similar to the experimental data. Accuracy of the inferred parameter values depends on the summary measures used. Here, we examined several sets of summary measures to determine a good set of the measures that can estimate the parameter values as accurate as possible. To this end, we prepared simulated sway data from the model with different sets of the parameter values, and examined the summary-measure dependent accuracy of the inference.