The validity of impact parameter estimation from the multiplicity of charged particles at low-intermediate energies is checked within the framework of ImQMD model. The simulations show that the multiplicity of charged particles cannot estimate the impact parameter of heavy ion collisions very well, especially for central collisions at the beam energies lower than $sim$70 MeV/u due to the large fluctuations of the multiplicity of charged particles. The simulation results for the central collisions defined by the charged particle multiplicity are compared to those by using impact parameter b=2 fm and it shows that the charge distribution for $^{112}$Sn +$^{112}$Sn at 50 MeV/u is different evidently for two cases; and the chosen isospin sensitive observable, the coalescence invariant single neutron to proton yield ratio, reduces less than 15% for neutron-rich systems $^{124,132}$Sn +$^{124}$Sn at $E_{beam}$=50 MeV/u, while the coalescence invariant double neutron to proton yield ratio does not have obvious difference. The sensitivity of the chosen isospin sensitive observables to effective mass splitting is studied for central collisions defined by the multiplicity of charged particles. Our results show that the sensitivity is enhanced for $^{132}$Sn+$^{124}$Sn relative to that for $^{124}$Sn+$^{124}$Sn, and this reaction system should be measured in future experiments to study the effective mass splitting by heavy ion collisions.