Cardiac electrophysiology mapping techniques now allow to record denser intra-operative electrograms (ECG). The patient-specific information extracted from these clinical recordings is extremely valuable. A growing field of research focuses on the personalization of electro-physiology models using this patient-specific information. The modeling in silico of a patient electrophysiology is needed to better understand the mechanism of cardiac arrhythmia. In the scope of ischemic cardiomyopa-thy, the predictive power of patient-specific simulations may also provide a substantial guidance in defining the optimal location of the implantable defibrillator, since all possible configurations could be tested in silico. This article describes an innovative personalization approach based on an unscented Kalman filter. Following an iterative process, the apparent conductivity is efficiently estimated in specific regions. A sensitivity analysis is performed to assess the filter parameters. With three patient cases, we finally demonstrate the accuracy and efficiency of our algorithm.