Ray,
When you say that something did not work, you are not giving much information for us to use to try to help you. Did the filter not work or did it not remove the drift?
How big is the drift, compared to the desired signal? How fast does it drift? How long are your sample sets? What are you doing with the ECG data? Looking at heart rate or analyzing a single-beat waveform for morphological changes? I think you can see that different end goals might require different approaches to baseline compensation.
If the drift is always fairly slow, you might be able to take small segments and subtract the mean of each segment. This may result in some offset or shift at the segment boundaries, but it is very simple and can easily be implemented without requiring a special toolkit.
Another approach is to low pass filter the ECG signal to extract an estimate of the baseline which can then be subtracted. Remember that all filters have a transient response and many will have a phase shift between the input and output.
Lynn
