Implementing FILTFILT using FILTER

Hello fellow Matlab users,

I am trying to understand how FILTFILT.m works & am trying to implement it using FILTER.m. To my understanding:

1. It uses reflected signal extension at the edges (left and right side of original signal).

2. Length of each reflected extension is 3*(filter_order-1).

3. Signal is filtered in forward direction.

4. Then filtered signal from 3 is filtered in backward direction.

5. Finally, filtered output from 4 (after flipping the filtered signal) corresponding to original signal length is chosen as output.

I have implemented the above in 2 m-files. One is a function I wrote (my_filtfilt.m) and the other one (compare_matlab_filtfilt_my_filtfilt.m) calls this function and Matlab filtfilt.m for comparison.

While results are close for a noisy Matlab sample ECG signal (>> load noisysignals x), they are not EXACTLY identical (see attached image/plot)!

Also attached are my m-files.

How to make the 2 outputs to match perfectly?

Any help would be greatly appreciated :)

Thanks!

Saif

Answer 1

[Martin Schätz]

Hi, my first idea is that the signal is not padded with zeroes, but mirrored (as is used in 2D convolution). Because as i understand filtfilt, you are doing all as it should be.