Date: 6/26/19 11:36 am From: Nagi Aboulenein <nagi.aboulenein...> Subject: Re: [Tweeters] Cordilleran Flycatcher and bird recordings
Hi Cara -
Just a note on bird song app recommendations. I'm primarily an Oregon birder, and we had Nathan Pieplow as the keynote speaker at our recent annual meeting. He stated that he recently switched from using SpectrumView to Song Sleuth, his newly preferred app for in-field bird song recording. Primarily he switched because Song Sleuth has a feature absent in other apps, namely a recording buffer.
Song Sleuth buffers up to 2 seconds of audio, and when you hit the record button it will record from the beginning of its buffer on-wards. I.e., it will start recording from the 2 seconds prior to when you actually hit the record button, which makes catching a bird song so much easier. With the other apps, you either invariably miss the beginning of the song, or you wind up hitting record and then hoping that the bird starts singing soon (which it usually won't :-) ).
Song Sleuth also has a song ID feature, but that currently misses more often than it gets it right (in fact some of its misses can be outright amusing). According to Nathan Pieplow, this isn't because the AI is bad - but rather it is because of lack of good recordings to train the AI. He stated that around 1000 good recordings are needed to train the AI on a particular species' songs and vocalizations, and very few birds have that many good-quality recordings. Even something as common as the Red-winged Blackbird, where there are a gazillion song recordings of the male, have almost no recordings whatsoever for the female. Contrast that to the Merlin app which is very good at getting bird IDs from photos, and that is simply because thousands and thousands of good photos exist for many species, covering males, females, juveniles, etc. In principle, the AI logic is the same for both apps - it is just a lack of the training data that's needed to train the AI for the bird sound apps.
All the best,
On Wed, Jun 26, 2019 at 11:07 AM Cara Borre <cmborre1...> wrote:
> Hello Tweets: > > We just returned from a great extended weekend birding in our beautiful SE > corner of the state with my goal of adding a Cordilleran Flycatcher to my > Washington list. The only way I was going to do this to my satisfaction > was to record the male's call and visualize the two part call on a > spectrogram after uploading it to ebird. Our more common Pacific-slope > Flycatcher has a similar call but it is one up-slurred note. This is hard > for the ear to detect, but easy to see on a spectrogram. > > I realize there is controversy regarding whether these birds should have > been separated or if they should still be called "Western Flycatcher", the > AOU will sort that out, but my quest was to merely add Cordilleran if I > could feel good about my documentation. I recorded all Pac-slope types I > heard calling and could call none Cordilleran. I would encourage recording > calling birds when trying to document Cordilleran. Here are some links and > tools to help: > > 1. Listen to Nathan Pieplow at the following podcast. I will be > purchasing his book the Peterson Field Guide to Bird Sounds of Western (or > Eastern) North America. He explains spectrograms in detail and why we > should record bird songs in the field and upload them to ebird. > > > https://podcasts.apple.com/us/podcast/bcr-166-nathan-pieplow-peterson-bird-sounds-of-western-na/id1128139090?i=1000434484164 > > 2. Read Nathan's article on "The Western Flycatcher problem" on his > Earbirding blog: > > http://earbirding.com/blog/archives/2996 > > 3. If you can figure out how to use the app he recommends for recording > in the field, it's called SpectrumView (for iphone); please teach me how, I > found it somewhat confusing. I prefer Voice Record Pro to record in the > field with my iphone I then email the recording to myself and upload the > file to ebird as you would a photo. Ebird generates the spectrogram and it > is here that I visually confirmed every "Western Flycatcher" I heard and > recorded was Pacific-slope. > > Analyzing spectrograms is also a way to potentially ID an unknown bird > vocalization using Nathan's book and I'm sure reliable phone apps in the > future. Ebird really wants to increase it's library of bird sounds as this > will be another way of documenting birds for those of us unable to master > the fine art of focusing a camera. That's why I shoot video, sooner or > later something's in focus! > > Hope you find this information as exciting as I did. > > As an aside, anyone wanting to go on a hunt for Kentucky Warbler? I have > the next few days off (and my birding buddy does not), please contact me. > > Cheers, > > Cara Borre > Gig Harbor > _______________________________________________ > Tweeters mailing list > <Tweeters...> > http://mailman11.u.washington.edu/mailman/listinfo/tweeters >