[en] Soft inelastic QCD processes are the dominant proton-proton interaction type at the LHC. More than 20 of such collisions pile up within a single bunch-crossing at ATLAS, when the LHC is operated at design luminosity of L=10³⁴ cm^-2s^-1 colliding proton bunches with an energy of √(s)=14 TeV. Inelastic interactions are characterised by a small transverse momentum transfer and can only be approximated by phenomenological models that need experimental data as input. The initial phase of LHC beam operation in 2009, with luminosities ranging from L=10²⁷ to 10³¹ cm^-2s^-1, offered an ideal period to select single proton-proton interactions and study general aspects of their properties. As first part of this thesis, a Minimum Bias trigger was developed and used for data-taking in ATLAS. This trigger, mbSpTrk, processes signals of the silicon tracking detectors of ATLAS and was designed to fulfill efficiently reject empty events, while possible biases in the selection of proton-proton collisions is reduced to a minimum. The trigger is flexible enough to cope also with changing background conditions allowing to retain low-p_T events while machine background is highly suppressed. As second part, measurements of inelastic charged particles were performed in two phase-space regions. Centrally produced charged particles were considered with a pseudorapidity smaller than 0.8 and a transverse momentum of at least 0.5 or 1 GeV. Four characteristic distributions were measured at two centre-of-mass energies of √(s)=0.9 and 7 TeV. The results are presented with minimal model dependency to compare them to predictions of different Monte Carlo models for soft particle production. This analysis represents also the ATLAS contribution for the first common LHC analysis to which the ATLAS, CMS and ALICE collaborations agreed. The pseudorapidity distributions for both energies and phase-space regions are compared to the respective results of ALICE and CMS.