Membrane fusion of the viral and host cell membranes is the initial step of virus infection and is catalyzed by fusion peptides. Although the beta-sheet structure of fusion peptides has been proposed to be the most important fusion-active conformation, it is still very challenging to experimentally identify different types of beta-sheet structures at the cell membrane surface in situ and in real time. In this work, we demonstrate that the interface-sensitive amide II spectral signals of protein backbones, generated by the sum frequency generation vibrational spectroscopy, provide a sensitive probe for directly capturing the formation of oligomeric beta-sheet structure of fusion peptides. Using human immunodeficiency virus (HIV) glycoprotein GP41 fusing peptide (FP23) as the model, we find that formation speed of oligomeric beta-sheet structure depends on lipid unsaturation. The unsaturated lipid such as POPG can accelerate formation of oligomeric beta-sheet structure of FP23. The beta-sheet structure is more deeply inserted into the hydrophobic region of the POPG bilayer than the a-helical segment. This work will pave the way for future researches on capturing intermediate structures during membrane fusion processes and revealing the fusion mechanism.