More algorithms for Generic Acoustic Components #297

towsey · 2020-02-19T03:51:53Z

I will finish off the Stacked Harmonics algorithm and create 2 or 3 new ones, e.g. a positive-sloped ridge and a negative-sloped ridge algorithm and a spectral-peak-track algorithm. The horizontal ridge algorithm would be pretty much equivalent to the existing whistle recognizer. There are possibly other component recognizers worth trying.

I will create a new branch in which all of these will be included.

Issue #297 Set up artificial spectrogram containing stacked harmojnics with various formant spacings. THis is to create a fixed test for unit testing.

Issue #297 Write unit test for recognising harmonics using an artificial spectrogram.

Issue #297 Make changes to the harmonic recognizer and rework its unit test and also test on recrding of Curlew calls.

Issue #297 THis also deals with Issue #307 Restore line that calculates the overlap given the provided framestep. THis must be done because code depends on the overlap rather than the framestep.

Issue #297 REmove comment line. THis class is now finished. It detects stacked harmonics in a test file. It also detects harmonics in a recording of curlew calls and human speech. The results on real data depend on the qulaity of the recorded calls/speech.

Issue #297 Recognizer substantially completed. Failing unit test in one special case which needs more work. Is p[icking up curlew spectral tracks.

Issue #297 Tweak the spectral peak track recognizer to get it workin on unit test spectrogram, boobook owl and the stone curlew.

Issue #297 Store information about track in the acoustic event.

Issue #297 Include ENUM of event types in EventBase class.

Issue #297 Set up a CLICK event type.

Issue #297 Add in option for user to combine acoustic events resulting from all profiles.

Issue #297 Develop the Click recognizer. Set up a test and begin debugging the click recognition algorithm

Issue #297 Finish unit tests for click events and also add in the possibility of detecting vertical ridges on real recordings.

Issue #297 THis method is to enable a series of rapid narrowband clicks (typical of some bird calls) to be combined into one event.

Issue #297 Write new recognizer for VerticalTrack components. THis was previously going to be called a Whip recognizer.

Issue #297 Remove contents of old class and repurpose for keeping path of a spectral track. The old class was not being used. The new class is now used by VerticalTrackParameters and will be used by SpectralPeakTrackParameters. Changes in Image_MultiTrack.cs are resharper inspired. But this class was the only one to use the old SpectralTracks.

Issue #297 Small changes to comments etc

Issue #297 Get new VerticalTrack component recognizer working. Get unit test working on existing artificial spectrogram test. It should give identical result to Click recognizer but currently is dropping the detected event band by one frequency bin.

Issue #297

Issue #297 Proper search for vertical tracks requires that outer loop is over freq bins and inner loop is over frames.

Issue #297 Use the SpectralTrack class for storing information about horizontal and vertical tracks.

Issue #297 Change one method and adjust the AcousticEvent and other classes to accomodate. Add parameter "CombineOverlappedevents" to the GenericRecognizer.yml file.

Issue #297 Enable the user to determine whether certain kinds of events are to be combined.

Issue #297 Set up different track drawing method for each type of spectral track.

Issue #297 THis is last commit before starting with the new event hierarchy.

Issue #297 Augment the UnitConverters class to make suitable for cases where do not use image info.

Issue #297 esp. drawing of composite events.

Issue #297

Issue #297 Fixed the CombineEvents method to prevent a kind of recursion! This problem arose due to having multiple kinds of Event class.

Issue #297 Used the Interval class to set the possible score range of an event.

Issue #297

Issue #297 Start work on Boobook Owl recognizer.

Issue #297 THis boobook recognizer is the first one which is species specific but extracts generic components and then does post-processing on the find primitive components.

Issue #297 Add method to draw tracks on spectrgram.

Issue #297 Also add in some debug code t o view tracks directly on spectrgrams.

Issue #297

Issue #297 Boobook Owl

Issue #297 Shift frog yml files to own folder

Issue #297

Issue #297 Start Koala recognizer version 3.

Issue #297

Issue #297 Set up the post-processing steps. Need to check if we can avoid all the casting up and down.

Issue #297 Set up everything except the post-processing.

Issue #297 To contain methods that help with dealing with lists of mixed event types.

Issue #297 Anthony works on casting between event types.

Issue #297

Issue #297 This is way of filtering out false positive whistles.

Issue #297

Issue #297 Ensure that the normalised score of an event does not exceed 1.0;

Issue #297

Issue #297: also small edits to Botaurus recognizer.

Issue #297

Issue #297 branch for recognizers.

atruskie changed the title ~~More recognizers for Generic Acoustic Components~~ More algorithms for Generic Acoustic Components Feb 19, 2020

atruskie assigned towsey Feb 19, 2020

atruskie added the enhancement label Feb 19, 2020

towsey mentioned this issue Mar 3, 2020

Fix display of Debug spectrograms and write unit tests #300

Closed

towsey added a commit that referenced this issue Mar 21, 2020

Set up artificial spectrogram

767dd01

Issue #297 Set up artificial spectrogram containing stacked harmojnics with various formant spacings. THis is to create a fixed test for unit testing.

towsey added a commit that referenced this issue Mar 22, 2020

Establish unit test for generic recognition of harmonics

ad7b7dc

Issue #297 Write unit test for recognising harmonics using an artificial spectrogram.

towsey added a commit that referenced this issue Mar 23, 2020

Work on the harmonic recognizer

2b9fa60

Issue #297 Make changes to the harmonic recognizer and rework its unit test and also test on recrding of Curlew calls.

towsey added a commit that referenced this issue Mar 24, 2020

Find bug re window overlap

4a3916e

Issue #297 THis also deals with Issue #307 Restore line that calculates the overlap given the provided framestep. THis must be done because code depends on the overlap rather than the framestep.

towsey added a commit that referenced this issue Mar 25, 2020

Work on spectral peak track recognizer

7ea052d

Issue #297 Recognizer substantially completed. Failing unit test in one special case which needs more work. Is p[icking up curlew spectral tracks.

towsey added a commit that referenced this issue Mar 26, 2020

Issue#297 Tweak the spectral peak track recognizer

b603703

Issue #297 Tweak the spectral peak track recognizer to get it workin on unit test spectrogram, boobook owl and the stone curlew.

towsey added a commit that referenced this issue Mar 27, 2020

Retain track info to store in acoustic event

5ae9cc5

Issue #297 Store information about track in the acoustic event.

towsey added a commit that referenced this issue Mar 27, 2020

Define a set of event types

17bf9f4

Issue #297 Include ENUM of event types in EventBase class.

towsey added a commit that referenced this issue Mar 27, 2020

Set up new event type - the Click

cc44224

Issue #297 Set up a CLICK event type.

towsey added a commit that referenced this issue Mar 27, 2020

Update GenericRecognizer.cs

a81b906

Issue #297 Add in option for user to combine acoustic events resulting from all profiles.

towsey added a commit that referenced this issue Mar 30, 2020

Write and Test Click recognizer

f941679

Issue #297 Develop the Click recognizer. Set up a test and begin debugging the click recognition algorithm

towsey added a commit that referenced this issue Mar 31, 2020

Finish unit test for click events

a43134a

Issue #297 Finish unit tests for click events and also add in the possibility of detecting vertical ridges on real recordings.

towsey added a commit that referenced this issue Mar 31, 2020

Write method to combine proximal similar events

2658be2

Issue #297 THis method is to enable a series of rapid narrowband clicks (typical of some bird calls) to be combined into one event.

towsey added a commit that referenced this issue Apr 1, 2020

Write new recognizer for VerticalTrack components

d5bf5d7

Issue #297 Write new recognizer for VerticalTrack components. THis was previously going to be called a Whip recognizer.

towsey added a commit that referenced this issue Apr 2, 2020

Update SpectralPeakTrackParameters.cs

4a16758

Issue #297 Small changes to comments etc

towsey added a commit that referenced this issue Apr 2, 2020

Fix out of range bugs

6f2c616

Issue #297

towsey added a commit that referenced this issue Apr 2, 2020

Update VerticalTrackParameters.cs

3c55a34

Issue #297 Proper search for vertical tracks requires that outer loop is over freq bins and inner loop is over frames.

towsey added a commit that referenced this issue Apr 3, 2020

Fix bugs in SpectralTrack class

82f78cd

Issue #297 Use the SpectralTrack class for storing information about horizontal and vertical tracks.

towsey added a commit that referenced this issue Apr 4, 2020

Change SpectralTrack method to return complete track data

50a4914

Issue #297 Change one method and adjust the AcousticEvent and other classes to accomodate. Add parameter "CombineOverlappedevents" to the GenericRecognizer.yml file.

towsey added a commit that referenced this issue Apr 4, 2020

Enable user to specify whether to combine events

5db59e7

Issue #297 Enable the user to determine whether certain kinds of events are to be combined.

towsey added a commit that referenced this issue Apr 7, 2020

Set up new methods for drawing spectral tracks

297729e

Issue #297 Set up different track drawing method for each type of spectral track.

towsey added a commit that referenced this issue Apr 9, 2020

Set up methods to draw tracks and acoustic events on spectrograms.

420e033

Issue #297 THis is last commit before starting with the new event hierarchy.

towsey added a commit that referenced this issue Apr 12, 2020

Augment unit converters class

117994e

Issue #297 Augment the UnitConverters class to make suitable for cases where do not use image info.

towsey added a commit that referenced this issue Apr 28, 2020

Get unit tests working for spectral events

76573b7

Issue #297 esp. drawing of composite events.

towsey added a commit that referenced this issue Apr 29, 2020

Fixed the drawing of labels and score bars for spectral events.

78572a8

Issue #297

towsey added a commit that referenced this issue Apr 29, 2020

Update CompositeEvent.cs

d69099c

Issue #297 Fixed the CombineEvents method to prevent a kind of recursion! This problem arose due to having multiple kinds of Event class.

towsey added a commit that referenced this issue Apr 29, 2020

Reworked calculation of normalised score

5afbf11

Issue #297 Used the Interval class to set the possible score range of an event.

towsey mentioned this issue Apr 29, 2020

Issue297 continued #316

Merged

7 tasks

towsey added a commit that referenced this issue May 1, 2020

Fix various bugs in the drawing of events on spectrograms.

65a980c

Issue #297

towsey added a commit that referenced this issue May 1, 2020

Create NinoxBoobook.cs

034d396

Issue #297 Start work on Boobook Owl recognizer.

towsey added a commit that referenced this issue May 1, 2020

Start work on Boobook recognizer

66d41d0

Issue #297 THis boobook recognizer is the first one which is species specific but extracts generic components and then does post-processing on the find primitive components.

towsey added a commit that referenced this issue May 2, 2020

Update Track.cs

925e2ca

Issue #297 Add method to draw tracks on spectrgram.

towsey added a commit that referenced this issue May 2, 2020

Set up the boobook owl recognizer.

b5e7169

Issue #297 Also add in some debug code t o view tracks directly on spectrgrams.

towsey added a commit that referenced this issue May 2, 2020

Update PezoporusOccidentalis.cs

d556710

Issue #297

towsey added a commit that referenced this issue May 2, 2020

Update NinoxBoobook.cs

24e560c

Issue #297 Boobook Owl

towsey added a commit that referenced this issue May 2, 2020

Put frog yml files in frog folder

4f48746

Issue #297 Shift frog yml files to own folder

towsey added a commit that referenced this issue May 2, 2020

Shift more frog yml files

c90f8fc

Issue #297

towsey added a commit that referenced this issue May 2, 2020

Create KoalaV3.cs

aec2584

Issue #297 Start Koala recognizer version 3.

towsey added a commit that referenced this issue May 2, 2020

Koala yml file

279c487

Issue #297

towsey added a commit that referenced this issue May 4, 2020

Finish work on Boobook owl recognizer

b44b7dd

Issue #297 Set up the post-processing steps. Need to check if we can avoid all the casting up and down.

towsey added a commit that referenced this issue May 4, 2020

Set up the Australasian Bittern recognizer

a11b60c

Issue #297 Set up everything except the post-processing.

towsey added a commit that referenced this issue May 5, 2020

Wirte new class EventExtensions.cs

e1ea09d

Issue #297 To contain methods that help with dealing with lists of mixed event types.

towsey added a commit that referenced this issue May 5, 2020

Change type casting between event types

5d2e8cc

Issue #297 Anthony works on casting between event types.

towsey added a commit that referenced this issue May 5, 2020

Final tweaks to Aust Bittern recognizer

d929e75

Issue #297

towsey added a commit that referenced this issue May 6, 2020

Work on reducing false positives in whistle recognizers.

5fa7cfb

Issue #297

towsey added a commit that referenced this issue May 7, 2020

Write methods that detect acoustic activity in band above whistle

04c95d5

Issue #297 This is way of filtering out false positive whistles.

towsey added a commit that referenced this issue May 7, 2020

Fix methods for post-processing of whistle events

b8880db

Issue #297

towsey added a commit that referenced this issue May 7, 2020

Update EventCommon.cs

bf2bd6f

Issue #297 Ensure that the normalised score of an event does not exceed 1.0;

towsey added a commit that referenced this issue May 7, 2020

tweak bittern recognizer

e52411a

Issue #297

towsey added a commit that referenced this issue May 7, 2020

Prepare config file for Koala recognizer v3

8ec397c

Issue #297: also small edits to Botaurus recognizer.

towsey added a commit that referenced this issue May 8, 2020

Changes as part of code review.

deee017

Issue #297

atruskie closed this as completed in #316 May 8, 2020

towsey added a commit that referenced this issue May 11, 2020

Final edits after Anthony review.

6f4eda6

Issue #297 branch for recognizers.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

More algorithms for Generic Acoustic Components #297

More algorithms for Generic Acoustic Components #297

towsey commented Feb 19, 2020 •

edited by atruskie

Loading

More algorithms for Generic Acoustic Components #297

More algorithms for Generic Acoustic Components #297

Comments

towsey commented Feb 19, 2020 • edited by atruskie Loading

towsey commented Feb 19, 2020 •

edited by atruskie

Loading