中国物理学会高能物理分会第十三届全国粒子物理学术会议（2021）

Liquid scintillator detectors play an important role in neutrino experiments, for its low threshold and high resolution in energy measurements. Traditionally they are treated as event calorimeters, which is not enough for high precision measurements. Low energy neutrino events may pile-up with radioactive background such as ${}^{14}$C, smearing the energy spectrum if pile-ups are treated as one. In this presentation we will introduce a general, extensible statistical reconstruction method for dual point source. By constructing a more precise detector response model and applying Bayesian analysis with suitable approximation, it can extract all information from PMT waveforms. Since it contains no black-box compared to deep learning algorithms, the results have clear statistical meaning and reliability.