A method of processing a raw acceleration signal, measured by an accelerometer-based compression monitor, to produce an accurate and precise estimated actual depth of chest compressions. The raw acceleration signal is filtered during integration and then a moving average of past starting points estimates the actual current starting point. An estimated actual peak of the compression is then determined in a similar fashion. The estimated actual starting point is subtracted from the estimated actual peak to calculate the estimated actual depth of chest compressions. In addition, one or more reference sensors (such as an ECG noise sensor) may be used to help establish the starting points of compressions. The reference sensors may be used, either alone or in combination with other signal processing techniques, to enhance the accuracy and precision of the estimated actual depth of compressions.