Skip to content

How It Works

Kait Frasier edited this page Jan 16, 2020 · 44 revisions

Typical Workflow

  1. Flow Chart
  2. Main Functions and Files Description
  3. Initial Data Preparation
  4. User Data Settings
  5. Species Data Settings
  6. DetEdit Interface
  7. Data Analysis

Flow Chart

The following diagram shows the processing pipeline to get all the parameters and run the interface:

FlowChart

Main Functions and Files Description

Function Name Description
detEdit Launches the interface and returns user decisions in files grouping the detection labels in different annotation MAT files (TD – true detections, FD – false detections, ID – detection types
myDataSettings Users’ specific parameters to run detEdit, mkLTSAsessions, modDet, and summaryParams. Parameters can be specified to overwrite default parameters and species default parameters.
Edetect, Edetect_wav Simple energy threshold detector to detect signals from audio files in wave file format (.wav; Edetect_wav) or extended wave file format (.x.wav; Edetect).
mkLTSAsessions Makes LTSAs per bout from LTSA of the corresponding deployment. Returns vector containing start times of spectral averages and power spectral densities into *LTSA.mat files.
make_TPWS Groups detections in bouts and returns click parameters into *TPWS.mat files.
modDet Deletes false detections stored in *FD.mat files from *TPWS.mat files and generates new *TPWS2.mat files. Optionally deletes ID detections stored in *ID.mat files as well.
initSpParams Sets up visualization parameters for specific species. Parameters for eleven species of odontocetes are predefined.
summaryParams Plots peak-to-peak amplitude, peak frequency, inter-click-interval histograms, and daily and weekly presence.
File Name Description
TPWS.mat Parameters of detected signals. Vector of detections start times (MTT), vector of peak-to-peak received sound pressure levels (MPP), matrix of time series (MSN), and matrix of power spectra (MSP).
LTSA.mat LTSA parameters per bout. Matrix of binned power spectral densities (pwr) and vector of start times associated with each bin (pt).
.ltsa File compression for audio data based on averaging spectra over long-term periods
TD.mat Vector of true detection times
FD.mat Vector of false detection times
ID.mat Matrix of ID detection classified based on color ID color code. 1st column contains detection times and 2nd column the ID label

Initial data preparation

Create parameter files required to use the interface:

  1. Make LTSA files (.ltsa)
  2. Make TPWS files (TPWS.mat)
  3. Make LTSA snippets files (LTSA.mat)

1. Make Long-Term Spectral Average (LTSA) files

LTSA files (.ltsa) are generated from a collection of WAV or XWAV files by averaging spectra over long-time periods and arranging these spectra sequentially as frequency-time spectrogram plots. LTSA files provide a quick overview of long-term recordings.

Using the Command Window in MATLAB, the .ltsafile is created as follows:

>> mkLtsa

and follow the prompt windows to specify the audio file format, file directories, and parameters (e.g., time average length and frequency bin size to average the data).

Only 5 filename formats of WAV and XWAV files are supported:

  • yymmdd-HHMMSS : SiteName_190326-123700.wav
  • yymmdd_HHMMSS : SiteName_190326_123700.wav
  • yyyymmdd_HHMMSS: SiteName_20190326_123700.wav
  • yymmddHHMMSS : SiteName_190326123700.wav
  • yyyymmddTHHMMS : SiteName_20190326T123700.wav

2. Make TPWS files (start Time, Peak-to-peak amplitude, Waveform, and Spectra parameters)

A *_TPWS.mat file contains the following matrices of detected signal parameters:

Variable Description
MTT Vector of start times of detections
MPP Vector of received level amplitudes (dBpp)
MSP Matrix of detection spectra, where columns are dictated by the size of the fast-Fourier transform (FFT) used to generate the spectra each row represents a unique detection.
MSN Matrix of waveforms, where columns are dictated by the window length used to extract the time series and each row represents a unique detection.

Create *_TPWS.mat file is required to call the interface, so the user can:

  • provide start times of detected acoustic signals to create the *_TPWS.mat file as follows:

    >> edit make_TPWS

    In the editor window, modify input/output locations and detection parameters to run the script.

  • apply a generic detector if you do not have any acoustic detections:

    • for use with XWAV files:

          >> Edetect

      The user is prompted to supply inputs including a text file or spreadsheet containing detection parameters (see GOM_BW_pfile_320.xlsx for example), and a directory containing XWAV files.

    • for use with WAV files:

          >> Edetect_wav ('paramFile','E:\DetEdit\GOM_BW_pfile_320.xlsx',...
          'tfFile','E:\MyTransferFunctionFiles\example_sig1_invSensit.tf',...
          'timeFile','E:\DetEdit\BW_Effort.xlsx',...
          'wavDir','E:\MyWAVFiles',...
          'channel',1)

3. Make LTSA snippets

After *_TPWS.mat files have been created, the following step is making the snippets of LTSAs corresponding to bouts of detections saved in TPWS files.

A *_LTSA.mat file contains the following matrices of detected signal parameters:

Variable Description
pt Vector of start times of spectral averages
pwr Matrix of power spectral density averages, where columns are dictated by the step size used to average frequency spectrum bins.

Using a script containing the user data settings (see myDataSettings for example), the LTSA snippets for each detection bout is performed as follows:

>> mkLTSAsessions(@myDataSettings)

or you can invoke the function directly

>> mkLTSAsessions

and it will prompt you with a window to select the data settings script.

Assumming mkLTSAsessions invocation, the data settings scripts should contain the required following variables:

User Data Settings Script

Before invoking any of the analysis (mkLTSAsessions, detEdit, modDet, or summaryParams), create your data settings script to define directories and parameters. Examples of data settings scripts are found in DetEdit\Settings folder. You can make different versions of these templates, with different names for different species, sites or analysis.

Variable Description mkLTSAsessions detEdit modDet summaryParams
filePrefix TPWS.mat files name to match :heavy_check_mark: :heavy_check_mark: :heavy_check_mark: :heavy_check_mark:
iterationNum Iteration number :heavy_check_mark: :heavy_check_mark: :heavy_check_mark: :heavy_check_mark:
sampleRate Your data sample rate :heavy_check_mark: :heavy_check_mark: :heavy_check_mark: :heavy_check_mark:
sp Species code :heavy_check_mark: :heavy_check_mark: :heavy_check_mark: :heavy_check_mark:
tpwsDir TPWS.mat files directory :heavy_check_mark: :heavy_check_mark: :heavy_check_mark: :heavy_check_mark:
tfName Transfer function file (.tf) directory :large_orange_diamond: :large_orange_diamond: :large_orange_diamond: :large_orange_diamond:
ltsaDir LTSA.mat files directory :heavy_check_mark:

:heavy_check_mark: = required variable :large_orange_diamond: = optional variable (only if transfer function was not previously applied)

More parameter preferences can be specified to override default parameters (see myDataSettings for more details). Uncomment the parameters as needed to overwrite the default settings specified in initDefaultParams.

Species Data Settings Script

DetEdit is provided with initSpParams with signal-specific parameters for the 4 different analysis (mkLTSAsessions, detEdit, modDet, and summaryParams). Different odontocete species parameters and other signals are given in initSpParams

Species abbreviation code:

  • : Unknown signal
  • De : Delphinid
  • Pm : Physeter macrorhynchus - Sperm whale
  • Zc : Ziphius cavirostris - Cuvier’s beaked whale
  • Me : Mesoplodon europaeus - Gervais' beaked whale
  • BWG: BWG signal type - Unidentified beaked whale
  • Md : BW31 signal type - Unidentified beaked whale
  • Ko : Kogia spp.
  • Po : Porpoise spp.
  • Dl : Delphinapterus leucas - Beluga whale
  • Mm : Monodon monoceros - Narwhal
  • MFA: Mid-frequency active sonar

If given a species code, the specific parameter preferences will override the default parameters (see parameters in initDefaultParams for more details). If you define different parameter values in myDataSettings, those will overwrite the species-specific parameters (initSpParams) and the default parameters (initDefaultParams).

DetEdit Interface

  1. Annotation and classification
  2. Editing tools

Using a script containing your data settings (see myDataSettings for example), you can invoke the interface as follows:

>> detEdit(@myDataSettings)

or you can invoke the interface directly

>> detEdit

and it will prompt you with a window to select the data settings script.

While running, the user will be presented with progress information and at the beginning will be asked to specify Starting session, use 1 to start with the first bout.

After, the interface panels will be displayed. Arrange the panel as you wish for better visualization:

Interface

1. Annotation and classification

Annotation files ( *_TD.mat, *_FD.mat, and *_ID.mat) are created if executing detEdit for the first time for a *_TPWS.mat file. All detections (stored in TPWS.mat file) are displayed in the interface. The annotations or classification you make would be stored in the corresponding annotation files, which will be updated each time a bout is edited.

Annotation File Description Color detections
*_TD.mat Vector of true detection times blue_dot
*_FD.mat Vector of false detection times red_dot
*_ID.mat Matrix of ID detection classified based on color ID color code. 1st column contains detection times and 2nd column the ID label

All annotation files from this stage are stored in the same directory where *_TPWS.mat files are located.

ID detection color legend:

ID Label RGB Color
1 color_type1
2 color_type2
3 color_type3
4 color_type4
5 color_type5
6 color_type6
7 color_type7
8 color_type8
9 color_type9
10 color_type10

A species label can be given to each ID value label. Default is given at initDefaultParams:

legend_species_code

You can modify the default species labels in myDataSettings in the following line:

line 77: paramsUser.mySpID = {'Label1','Label2','Label3', 'Label4', 'Label5', 'Label6', 'Label7', 'Label8', 'Label9', 'Label10'};

2. Editing tools and keyboard shortcuts

The annotation process can be done per bout by annotating the entire window at once or by selecting batches of detections using the Matlab's paintbrush tool - Brush_icon_default (yellow default color).

:warning: to activate the brush, click at the plot you want to brush the data and then press p. Brush button will appear selected and you will be ready to flag detections. You can select a different color for the brush which has different meanings.

Keyboard shortcuts:

Keyboard Brush color Description
t All detections in current window as true
f All detections in current window as false
u Update window contents
space Jump to next consecutive bout (next window session)
b Jump back one bout (previous window session)
Brush_icon_default + r Brush_icon_red + u Detections selected as false
Brush_icon_default + g Brush_icon_green + u Detections selected as true
Brush_icon_default + y Highlights and displays summary info in black outlined (black_circle) of selected clicks (:warning: but does not change annotation files)
Brush_icon_default + ID Label (e.g. Brush_icon_default + 1) ID_labeling_color_code + u Detections selected as ID color code

Other keyboard shortcuts to modify the interface settings, this will prompt the user to introduce the desired number:

Keyboard Description
j Jump to a non-consecutive bout
a Adjust LTSA contrast and brightness
s Adjust maximum time between detection scale
h Adjust peak frequency scale
d Adjust received levels (dBpp ) scale
m Adjust transformed received levels (dBrms ) scale
'<' Change transformed received levels (dBrms ) threshold
: Change received levels (dBpp ) threshold
^ Change peak frequency threshold
e Re-label selected data
If points are brushed, the new label will be applied to the set of selected points with the chosen ID number. If no points are brushed, then all points in the current window with the selected ID will be re-labeled.

Data Analysis

  1. False Positive Rate Evaluation
  2. Modify Annotation Files
  3. Summarize Parameters

1. False Positive Rate Evaluation

To estimate the dataset's false positive rate, the following keyboard shortcuts will set the interface to display random detections for evaluation:

  • x: evaluation at the signal level, where random individual detections are displayed.
  • w: evaluation at the time-bin interval level, where a batch of detections within the defined interval length is displayed and the user has to evaluate if at least one detection within the interval is true.

Then, the evaluation decision is done by pressing the following keyboard shortcuts:

  • x: as true detection
  • z: as false detection

The random detections are selected systematically based on a given step size. The user can specify this parameter at myDataSettings, by uncommenting the line as follows:

line 42: paramsUser.c4fd = your number;

this will overwrite the default step size parameter (see initSpParams for more details of each species step size). If no species name is specified the step size parameter is 1, where every single click is evaluated one by one (see initDefaultParams).

All evaluation decisions are stored in the *_TD.mat files.

2. Modify Annotation Files

After manually editing the detections using the interface, the user can remove the annotated detections as false positive detections stored in FD.mat file from the *_TPWS.mat files as follows:

>> modDet(@myDataSettings)

or you can invoke the function directly

>> modDet

and it will prompt you with a window to select the data settings script.

This will create TPWS2.mat files stored in a new folder (TPWS2 folder) within the current folder. The ID.mat files will be saved in the TPWS2 folder to maintain the ID detection labels.

Optionally, ID detections stored in ID.mat files can be excluded as follows:

line 71: paramsUser.excludeID = 1; % yes - 1 | no - 0. Exclude ID times from MTT files

or parsed out into sets using custom scripts.

Three different acoustic parameters, peak-to-peak received sound pressure levels, time-interval between detections, and peak frequencies from the detections of the modified file can be computed if desired as follows:

line paramsUser.calcParams = 1;% yes - 1 | no - 0. Calculate Parameters peak-to-peak RLs, inter-detection-interval and peak frequency 

The process of using detEdit and modDet can be repeated iteratively until all detections are annotated, or a sufficiently low percentage of false detections is obtained. For every iteration, new detection and annotation files are generated and stored together in a common directory to keep track of all changes.

3. Summarize Parameters

A data summary of all the data from one site or from all the files with the same filePrefix can be computed as follows:

>> summaryParams(@myDataSettings)

or you can invoke the function directly

>> summaryParams

and it will prompt you with a window to select the data settings script.

Histogram plots of peak-to-peak received sound pressure levels, time-interval between detections, and a time series of the weekly presence of the true detections at the signal level and the time-bin interval level will be displayed and stored in the corresponding directory.

summaryParams_plots

You would be required to specify an excel file (see Test dataset as an example) with the monitoring effort start and end times, and the reference time you want the time series plot to start. You can specify these parameters in myDataSettings in the following lines:

line 27: effortTimes = 'E:\Effort.xls'; % specify excel file with effort times
line 28: referenceTime = '2010-04-01'; %reference time format 'yyyy-MM-dd'
You can’t perform that action at this time.