GRACE and GRACE-FO space missions aimed to monitor the Earth�s gravity field for a plethora of Earth applications. For the determination of the Earth�s gravity field, the calibrated accelerometer measurements are crucial since they measure the non-gravitational accelerations which are subtracted from the satellite�s total accelerations. Thus, the accelerometers need to be calibrated, a rather formidable task. In this study, we propose an optimal, innovative calibration method based only on the measurements of the accelerometer and the total spacecraft accelerations determined from the GPS precise orbits (POD), without the use of any physical non-gravitational force model. The idea behind this method lies in the matched filter, an optimal method used extensively in radar applications for the detection of known pulses in scattered signals. The total satellite accelerations are calculated through a double numerical differentiation of the GPS positions and play the role of the calibrated transmitted signal to which the accelerometer signal needs to be matched and thus calibrated. Specifically, the penumbra transitions of each spacecraft to and from the Earth�s shadow, induce characteristic jumps in the measurements, which are ideal for the calibration of the accelerometer in reference to the POD. Following this concept, we retrieve daily accelerometer scale factors for both missions and estimate the daily biases through a least-squares process. Long-term scale factors along the three axes of the accelerometers show an interesting periodicity strongly correlated with the behaviour of the ?� angle, which determines the time a LEO satellite spends in direct sunlight.