search

MarineBioAcousticsRC / Triton

:whale: Scripps Whale Acoustics Lab :earth_americas: Scripps Acoustic Ecology Lab - Triton with remoras in development
Other
37 stars 27 forks source link

empty CSV output file when computing metrics on LTSA #91

Open chuetz opened 2 years ago

chuetz commented 2 years ago

Hi all,

I'm opening a new issue concerning my second problem. After managing to create a LTSA from my data (with the Soundscape remora), I tried to compute some metrics on it. I tried several parameters combinations, but the output CSV file is either empty or filled with NaN, with only the column headers correct.

I'm joining the log file in which the output of two different trials I made. I guess that I'm doing something wrong in setting the "Data analysis parameters", but I cannot figure out where.

My LTSA computation parameters were the following: -averaging time: 10 sec -frequency bin size: 200Hz

I therefore did put the following "Data analysis parameters": -bandpass edges: 20-10000 Hz (my wav data are sampled at 22kHz) -averaging: -bin size time (sec): 10 -PSD bin size frequency (Hz): 200 -Min. Seconds / average (%) : 1 or 50

Any suggestion would be greatly appreciated! Thanks, logfile.txt

sfregosi commented 2 years ago

Hi @chuetz thanks for providing all that info!

I was able to recreate an empty output csv like you describe, when I used a 'Bin Size Time' of 10 sec, 'PSD Bin Size Frequency' of 200, and 'Min. Seconds/Average' of 50%. I'm working with continuous data that is originally from xwavs but I don't think the file type should matter for this.

I actually think there are two issues to deal with.

(1) Frequency Bin Size I was able to successfully get data in my csv output when I either increased the bin size (e.g., 3600 sec), leaving 50% minimum seconds/avg, or decreasing that minimum (I had success at 1%), BUT I had to have the frequency bin size at 1 Hz to get any output for the PSD output. By definition, PSDs are in 1 Hz bands, so that needs to be set to 1 Hz to get any PSD output (regardless of time bin size or minimum seconds percent). I'm not sure if this is a bug or by default; I'll need to ask someone who knows the soundscape remora better about that and have them chime in. But, using 1 Hz should fix that problem.

(2) Time Bin Size/Minimum Seconds From the logtile.txt you uploaded, it looks like you had some success using 10 sec bin with 1% minimum, but not on all hours to be analyzed. I suspect that there are gaps in your data that is causing it to fail meeting that 1% minimum criteria on those hours. My thoughts here are based on the bit in your logfile.txt that says Time average 360/8871 computed; 27-Apr-2022 01:29:20; duration 0.0012381 s; 341 segments - for me it says 360 segments here, but at least the output looks like it is trying to do something here. I get the same log output if the frequency bin is 200 Hz, resulting in no output, but looking like things are trying to work.

Are your data duty cycled, or does it have any write gaps/missing periods, even if they are short? If so, I'd suggest you increase the bin size (try as big as 3600 seconds) and see if that works (with the frequency down to 1 Hz), then let me know!