Urban green open space is an important part of urban public space. Reasonable allocation of urban green open space is essential for the construction of a livable urban environment. Spatial accessibility is often used to evaluate the rationality of the spatial layout of public service facilities. Two-step floating catchment area method is an intuitive, simple and widely used accessibility calculation method. However, this method does not consider the differences among demand subjects. In this study, we focus on the central district of Nanjing, extract the user attributes and spatial distribution through mobile phone signaling data. Then we subdivide the demand subjects through classification and improve the search strategy. We ameliorate the disadvantages of the traditional study of the two-step floating catchment area method that does not consider the variability of different groups, statistical data lags, and coarse spatial scales. The results show that the accessibility of green open space shows obviously spatial differentiation. The accessibility of Jiangnan area is generally higher than that of Jiangbei new area. The low value areas are concentrated around the central city with high population density and small scale of green open space resources. The overall pattern of green open space accessibility is similar for different groups. However, in terms of values, the mean accessibility of the elderly is 14.60% of that of the other age groups. The working population is 86.02% of that of the residential population. As a disadvantaged group, the elderly people are marginalized in the access to green open space services. The results can provide a scientific basis for optimizing the layout of green open space in Nanjing, and can also provide reference for the accessibility evaluation of other public service facilities.