An improved time domain method (ITDM) is proposed for moving force identification using bridge responses, which aims at an acceptable solution to the ill-conditioning problem that often exists in the inverse problem of moving force identification. Based on the method of moments (MOM) and the theory of moving force identification, the moving forces were described as a combination of whole basis functions, such as orthogonal Legendre polynomials or Fourier series, and were then estimated by solving the new system equations developed based on the basis functions. Under a number of response combination cases, the moving vehicle loads are identified using the ITDM and compared with the existing time domain method (TDM). Further a laboratory study was conducted to evaluate the effect of various parameters on the ITDM. Those parameters include basis function number, mode number, number of measured stations, and CPU executive time of the ITDM. Simulation and experiment results show that the ITDM has higher identification accuracy and robust noise immunity as well as being able to generate an acceptable solution to the ill-conditioning problem to some extent when it is used to identify the moving forces from bridge responses. Meanwhile, the ITDM can lessen the executive CPU time as well as being more flexible when compared with the TDM. This is beneficial to real time analysis of moving force identification in field.