Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

CatMeows: A Publicly-Available Dataset of Cat Vocalizations (2020)

zenodo.org

Post reply on HN
Print view View original

11–20 of 77 posts

Re: CatMeows: A Publicly-Available Dataset of Cat Vocalizations (2020)

#11

Post by gregsadetsky »

bingu wrote: post #8

Great and fun idea, but the resolution of these files is incredibly low.

You're not kidding, they're 8 kHz i.e. ""phone resolution""...

Interesting, this seems to be related to the Bluetooth microphone that they used:

"Thus, we adopted a very small and lightweight microphone placed under the cat’s throat through a collar"

The microphone is described in their paper [0]: "The signal is transmitted via Bluetooth [...] at a sampling frequency of 8 kHz."

Also: "As a consequence, the actual range of frequencies we could rely on is 0–4 kHz. We expected that the fundamental frequency emitted by cats falls within the range, whereas some higher-frequency harmonics were likely to be cut"

[0] https://www.mdpi.com/2076-2615/9/8/543/htm

Re: CatMeows: A Publicly-Available Dataset of Cat Vocalizations (2020)

#12

Post by orangecat »

adfm wrote: post #9

Black Mirror thought: Given the popularity of cats and that things like buzz pollination exist, it would be interesting to see if these sounds affect human behavior in some meaningful way.

https://www.sciencemag.org/news/2009/07/cat-purrs-evoke-baby...

Re: CatMeows: A Publicly-Available Dataset of Cat Vocalizations (2020)

#13

Post by nexuist »

bingu wrote: post #8

Great and fun idea, but the resolution of these files is incredibly low.

I'm very confused as to why. According to Finder these have an 8kHz sample rate? How did they find a mic that bad?

Or is it intentional? Is there some scientific usefulness that would make these clips easier for a machine to analyze? I really don't know if there would be, since a higher resolution clip could just be sampled again if you really wanted something lower.

It is very unfortunate, but I'm grateful that someone performed this experiment as ethically as they could think of and made the data available for all to use.

Re: CatMeows: A Publicly-Available Dataset of Cat Vocalizations (2020)

#15

Post by WalterBright »

ctoth wrote: post #10

This reminds me of some people working on a cat translator[0] The paper this dataset links to [1] seems to be using statistical techniques to compare spectrograms of meows, something that seems pretty easy with fastaudio[2]. [0]: https://github.com/FrogBoss74/RealCatTranslator [1]: https://www.mdpi.com/2076-2615/9/8/543 [2]: https://github.com/fastaudio/fastaudio

Having a cat, I have worked out a sophisticated translation lookup table:

    cat[] = meow meow meow meow meow
    translation[] = feedme feedme feedme feedme feedme
So far, it is 100% accurate.

Re: CatMeows: A Publicly-Available Dataset of Cat Vocalizations (2020)

#20

Post by tyingq »

nexuist wrote: post #13
bingu wrote: post #8

Great and fun idea, but the resolution of these files is incredibly low.

I'm very confused as to why. According to Finder these have an 8kHz sample rate? How did they find a mic that bad? Or is it intentional? Is there some scientific usefulness that would make these clips easier for a machine to analyze? I really don't know if there would be, since a higher resolution clip could just be sampled again if you really wanted something lower. It is very unfortunate, but I'm grateful that some…

Lol. It's a no-name bluetooth earbud from WalMart that's $16.

https://www.walmart.com/ip/QCY-Mini2-Wireless-Business-Headp...

Post reply on HN