I have a ECG signal measured as a test subject is performing 3 different activities - sitting, walking(2mph) and jogging(5mph) and also the corresponding 3-axis accelerometer(X,Y and Z) signals. I can visually observe correlation between the amplitude of accelerometer signal and noise in the ECG signal for the 3 activities. The following are my questions,
- How to decide on which axis of accelerometer data to use as noise feed to an adaptive LMS filter? Or should a combination of all 3 axes data be considered and if so, how?
- Does the accelerometer data have to be normalized to the amplitude range of the ECG signal for effective performance of LMS filter?
- What are the suggested parameters for the LMS filter for this case(The available literature is not clear on this problem).
- For implementation in a embedded platform, is RL, LMS or NLMS suggested?