A miniaturized accelerometer sensor attached to the heart may be applied for monitoring cardiac motion. Proper understanding of the sensor measurements is required for successful development of algorithms to process the signal and extract clinical information. In vivo testing of such sensors is limited by the invasive nature of the procedure. In this study we have developed a mathematical simulation model of an accelerometer attached to the heart so that testing initially may be performed on realistic, simulated measurements. Previously recorded cardiac motion by sonomicrometric crystals was used as input to the model. The three dimensional motion of a crystal attached to the heart served as the simulated motion of the accelerometer, providing the translational acceleration components. A component of gravity is also measured by the accelerometer and fused with the translational acceleration. The component of gravity along an accelerometer axis varies when the axis direction slightly rotates as the accelerometer moves during the cardiac cycle. This time-varying gravity component has substantial effects on the accelerometer measurements and was included in the simulation model by converting the motion to prolate spheroidal coordinates where the axis rotation could be found. The simulated accelerometer signal was filtered and integrated to velocity and displacement. The resulting simulated motion was consistent with previous accelerometer recordings during normal and ischemic conditions as well as for alterations of accelerometer orientation and patient positions. This suggests that the model could potentially be useful in future testing of algorithms to filter and process accelerometer measurements.