A novel and robust technique, vector representation (VR), was developed for classifying the glow curves of thermoluminescent dosimeters (TLDs) exposed to proton, photon, and mixed proton–photon fields produced by two types of TLDs:LiF:Mg,Ti (TLD-100) and CaF 2 :Tm (TLD-300). Vectors were constructed from glow curve points and classified based on the TL vector inner product with a unit vector and TL vector magnitude. The goal was to perform two-category (proton or mixed/photon) and three-category (proton, photon, or mixed) classification. The method was tested for accuracy using leave-one-out validation (LOOV) with classification based on the smallest Mahalanobis distance. Using the high-temperature peaks of each material, correct classification rates exceeding 92% in all cases were obtained.