Owens et al (2011) on their paper: "paraspinal muscle function assessed with the flexion relaxation ratio at baseline in a population of patients with back-related leg pain" suggested to use an inverse FRR to avoid the zero problem.

Cited:

Inverse FRR is

essentially the percentage to which the lumbar muscles

become electrically silent during full flexion in comparison

with the higher activity seen during forward flexion. Values

typically range between 0 and 1 because the fully flexed

(resting) EMG levels rarely exceed the EMG activity during

flexion. When 1/FRR is 0, it indicates that full electrical

silence in the minimum RMS EMG value was achieved; 1

would represent no silence at all in the minimum RMS

EMG value.

This is a possible solution but the direct FRR is commonly used, how do people avoid this problem?

Dear Costa,

as is well known, RMS of surfaceEMG (sEMG) is evualated by menas of the square root of the sum of the sEMG square samples x(n) taken in an epoch:

RMS-EMG = SUM(x(n)^^2)^^1/2, for n ranging in an epoch,

where typically:

epoch = 1 sec (for constant force - isometric contraction)

That is, epoch must be a time-interval of quasi-stationariety of the sEMG signal. In your question you say that, after filtering, the mean value is zero. What does it mean? Before filtering was non-zero? How is it possible that filtering can produce such effect? Moreover: what the mean value has to do with the RMS-EMG?

Finally, if you want the whatever coefficient C, evalued from two quantities A and B, varies between 0 and 1 you must  use the relation:

C = A / (A+B)

or, alternatively

C = B / (A+B).

I don't know if these comments will help you.  I stopped using the FRR data for diagnostic purposes due to the potential for errors leading to false positive findings.  Research ( T. Sihnovan, J. Partanen, O. Hanninen, Averaged (rms) surface EMG in testing back function, Electromyography and Clinical Neurophysiology 28 (1988), 335-339.) shows a test retest reliability of 0.93 for flexion using RMS values of peak amplitude and 0.98 for extension.  Coming up against gravity produces more consistent data than flexion as you are coming up against a constant force verus bending with it.  Given the documented reliability why use one which has questionable validity?   I have studied over 2000 low back pain sufferers following Sihnovan's data and found it to be very reliable.