[en] In this work the di-electron production in p+p and d+p reactions at a kinetic beam energy of 1.25 GeV/u measured by the HADES spectrometer is discussed. At E_kin=1.25 GeV/u, i.e. below the η meson production threshold in proton-proton reactions, the Δ Dalitz decay is expected to be the most abundant source above the π⁰ Dalitz decay region. The observed large difference in di-electron production in p+p and d+p collisions suggests that di-electron production in the d+p system is dominated by the n+p interaction. In order to separate Δ Dalitz decays and np bremsstrahlung the di-electron yield observed in p+p and n+p reactions, both measured at the same beam energy, has been compared. The main interest here is the investigation of iso-spin effects in baryonic resonance excitations and the off-shell production of vector mesons. We indeed observe a large difference in di-electron production in p+p and n+p reactions. Results of these studies will be compared to recent calculations. We will also present our experimentally defined cocktail for heavy-ion data. A strong excess of lepton pairs observed by recent high energy heavy-ion dilepton experiments hint to a strong influence of baryons, however no data exist at highly compressed baryonic matter, achievable in heavy-ion collisions from 8.45 GeV/u beam energy. These conditions would allow to study the expected restoration of chiral symmetry by measuring in-medium modifications of hadronic properties, an experimental program which is foreseen by the future CBM experiment at FAIR. The experimental challenge is to suppress the large physical background on the one hand and to provide a clean identification of electrons on the other hand. In this work, strategies to reduce the combinatorial background in electron pair measurements with the CBM detector are discussed. The main goal is to study the feasibility of effectively reducing combinatorial background with the currently foreseen experimental setup, which does not provide electron identification in front of the magnetic field. (orig.)