Skip to content
CSchoenbeck edited this page Sep 24, 2024 · 9 revisions

HOME > REMORAS > SOUNDSCAPE METRICS

Soundscape Metrics - developed for the NOAA National Marine Sanctuaries SanctSound Project

This Triton Remora calculates long-term spectral averages (LTSAs) from audio recordings with user defined time and frequency resolution. These LTSAs can then be read to compute soundscape metrics such as power spectral density, third-octave levels, octave levels, broadband levels, and hybrid millidecade over longer, user-specified time bins.

  1. Installation
  2. Remora: Soundscape Metrics

Installation

pic1

  • Start Triton
  • Select Remoras/ Add Remora - choose Soundscape Metrics to install
  • Triton needs to be restarted

REMORA: Soundscape Metrics

pic2

The pulldown menu for REMORA Soundscape Metrics has 4 options:

  • Options 1 and 2 are functioning for the purpose of the SanctSound Project with use of SoundTrap data; descriptions and caveats below
  • CAUTION:
    • Option “Load Soundscape LTSA” is for viewing purposes and only partially working
    • Option “Plot Soundscape Metrics” has not been implemented yet

Step 1: Make Soundscape LTSAs

General Settings

  • Define input and output directory, output filename by either selecting through the user interface or by typing into the text box

LTSA Parameter Settings

  • Define “Averaging Time” in seconds (default is 1 s)
  • Define “Frequency Bin Size” in Hertz (default is 1 Hz)
  • Define “Length of LTSA” in days (default is 3 days); however, using 7 days and 48 kHz sampling rate, file size per LTSA is ~57 GB; IMPROVEMENT: (maybe) all at once instead of increments
  • Define “Start with LTSA #”, in case one needed to resume computation mid deployment (default 1)
  • Define “File Type” (default 2 = XWAV); file type 1 = WAV; IMPROVEMENT: XWAV file type now works!
  • Define “Data Type” (default 1 = HARP); file type 5 = SoundTrap; IMPROVEMENT: HARP data type now works!
  • Define “Data Channel Number” (default 1); CAUTION: other channels not yet tested

Save/Load Settings

  • Optional "Save/Load Settings"; advisable to save settings for future documentation

pic3

Difference to "standard" LTSA

  • 32 versus 64 bit depth for byte location directory; smaller bit depth in “standard” LTSA when using higher compression factors (e.g. 5s / 100 Hz resolution for 100 kHz Nyquist data).
  • Integer versus floating point (single) values for spectral values.

Debugging Notes

  • Stepping into the Soundscape Metrics Remora code, sm_mk_ltsa.m is the primary code that makes the soundscape LTSAs. From this code a variety of other functions are called and used, but this is a good starting point to debug error messages.

Step 2: Compute Soundscape Metrics

General Settings

  • Define input and output directory by either selecting through the user interface or by typing into the text box
  • Define which input LTSA file number to start with in case only a subset of interest IMPROVEMENT: expand this to first/last file or (maybe) start/stop time
  • Select output file type; CSV or LTSA; multiple selections possible; CAUTION: output as LTSA not yet implemented – should be expanded to two output options – “standard” LTSA for e.g. ship detector use; “Soundscape Metrics” LTSA for further metrics computation

Data Analysis Parameter Settings

  • Define band edges of output product in Hertz, will crop spectra to desired low and high frequency band (default low: 20 Hz; default high: 24000 Hz)
  • Define averaging resolution in time and frequency (default 1 h, i.e. 3600 seconds; 1 Hz)
  • Define what percentage of coverage the average needs to have (e.g. default 3600 seconds average; 50% coverage means that at least 1800 seconds need to be available for the average to be computed); IMPROVEMENT: check which portion of the average is available and whether that is representative of the time period CAUTION: this setting doesn't always save when loading in saved parameter settings, so check this setting especially when loading in saved parameter settings
  • Select all analysis types to be computed; (default none selected, suggested selected all: Broadband Level, Octave Level, Third Octave Level, Power Spectral Density, Hybrid Millidecade)
  • Select averaging type to be computed; (default none selected, suggested: median and percentiles) – calculation based on arithmetic computation for mean, median and percentiles within the averaging time period / frequency bin IMPROVEMENT: add logarithmic mean calculation as a user option
  • When calculating Hybrid Millidecade, you must also select Power Spectral Density with the median averaging type because Hybrid Millidecade calculations use the Power Spectral Density.
  • Select removal of erroneous data; IMPROVEMENT: Remove Disk Writes HARP option available, other options not yet available

Data Calibration Settings

  • Select “Calibrate Data” if absolute received levels are desired
  • Choose between “Single Value” (for SoundTraps) and “Transfer Function” (for HARPs) calibration; IMPROVEMENT: no default selection of radio button
  • If “Single Value”: input full system sensitivity in decibels
  • If “Transfer Function”: choose transfer function either through the “Select TF” button and user interface or by typing the full path and file name into the text box

Save/Load Settings

  • Optional “Save/Load Settings”; advisable to save settings for future documentation; however, log file is also written when computation starts

cmpt_soundscape_image

Soundscape Metric Units

  • Power spectral density in dB re 1uPa^2/Hz
  • Broadband level in dB re 1uPa
  • Octave band level in dB re 1uPa
  • Third octave band level in dB re 1uPa
  • Hybrid millidecade in dB re 1uPa^2/Hz

Debugging Notes

  • Stepping into the Soundscape Metrics Remora code, sm_cmpt_metrics.m is the primary code that computes the soundscape metrics from the soundscape LTSAs created in step 1. This code calls and uses a variety of other functions, but is a good starting point to debug any errors.