Incorrect test measurements annotations count in mxm

[chesnokov-yuriy]

Dear George

Using -f flag in mxm I noticed some invalid output in the count of test measurements annotations variable ntmeas

Sincerely Yuriy

[0:00:01.025]        369     =     0     0     0 	71.76 
[0:00:01.872]        674     =     0     0     0 	70.82 
[0:00:02.722]        980     =     0     0     0 	70.59 
[0:00:03.566]       1284     =     0     0     0 	71.05 
[0:00:04.408]       1587     =     0     0     0 	71.29 
[0:00:05.219]       1879     =     0     0     0 	73.97 
[0:00:06.041]       2175     =     0     0     0 	72.97 
[0:00:06.875]       2475     =     0     0     0 	72.00 
[0:00:07.683]       2766     =     0     0     0 	74.23 
[0:00:08.483]       3054     =     0     0     0 	75.00 
[0:00:09.344]       3364     =     0     0     0 	69.68 
[0:00:10.205]       3674     =     0     0     0 	69.68 
[0:00:11.050]       3978     =     0     0     0 	71.05 
[0:00:11.894]       4282     =     0     0     0 	71.05 
[0:00:12.722]       4580     =     0     0     0 	72.48 
[0:00:13.563]       4883     =     0     0     0 	71.29 
[0:00:14.363]       5171     =     0     0     0 	75.00 
[0:00:15.177]       5464     =     0     0     0 	73.72 

measurements annotation saved in htr annotator

$ mxm -r 118e00 -a htr htr -f 0 -t 10
     Normalized RMS error:  0.0000% (n = 13)
Mean reference measurement: 66.3877

using -f 5 to start from 5th second results in the same n = 13 and incorrect mean reference measurement

$ mxm -r 118e00 -a htr htr -f 5 -t 10
     Normalized RMS error:  0.0000% (n = 13)
Mean reference measurement: 39.0408

It seems that despite the use of -f ntmeas is counted from the signal start resulting in altered RMS and mean reference measurement.

$ mxm -r 118e00 -a htr htr
     Normalized RMS error:  0.0000% (n = 2255)
Mean reference measurement: 65.2973

default -f is 5:00

$ mxm -r 118e00 -a htr htr -f 0
     Normalized RMS error:  0.0000% (n = 2255)
Mean reference measurement: 76.8367

starting from -f 0 gives exactly the same n = 2255 and different mean reference measurement

[chesnokov-yuriy]

as mxm is used for ANSI/AAMI-standard measurement-by-measurement annotation file comparator for FDA related tests using default mxm settings gives increased RMS performance and lower mean reference?