We are testing some new algorithms on P300 data that we previously recorded. However we had some weird problems and consequently we set up a simple scenario with a generic stream reader (that reads the P300 file) and an stimulation based epoch box, we added some debug output to the epoching box and this was the output
So the epocher was created around the time that the stimulus is received in the epoching box, i.e. perfect. A few cycles later at 9.93sec it processes its first chunk, but that chunk is timestamped at the time almost 250ms earlier. In an online scenario I could expect something like this when the processing runs late, but this is read from a file. Could it be that the file was not recorded well (although perfect classification was achieved on later files based on this training file) and that the chunks have not received the correct timestamp? What I would expect is the following where I replaced the file reader with a sinus oscillator:[ INF ] At time 9.906 sec <Box algorithm::Stimulation based epoching> Created new epocher for stimulus OVTK_StimulationId_Label_01 at time 9.898 sec:9.898 sec:10.898 sec
[ INF ] At time 9.930 sec <Box algorithm::Stimulation based epoching> Still processing epoch for stimulation OVTK_StimulationId_Label_01 with timestamps 9.898 sec:9.898 sec:10.898 sec
[ INF ] At time 9.930 sec <Box algorithm::Stimulation based epoching> Chunk start time: 9.656 sec, chunk end time: 9.688 sec
Here the chunk ranging from 9.875 to 10 sec is indeed processed not much later than 10sec. What causes the mismatch between the current time and the chunk times in the first scenario where one reads from the file?[ INF ] At time 9.898 sec <Box algorithm::Stimulation based epoching> Created new epocher for stimulus OVTK_StimulationId_Label_01 at time 9.898 sec:9.898 sec:10.898 sec
[ INF ] At time 10.008 sec <Box algorithm::Stimulation based epoching> Still processing epoch for stimulation OVTK_StimulationId_Label_01 with timestamps 9.898 sec:9.898 sec:10.898 sec
[ INF ] At time 10.008 sec <Box algorithm::Stimulation based epoching> Chunk start time: 9.875 sec, chunk end time: 10.000 sec
Best regards,
Dieter