الجمعة، 11 مارس 2022

Some Pig: Scientists Use AI To Decode Swine Sounds

(Photo: Steven Weeks/Unsplash)
If you watched the Disney heartwarmer Up and found yourself wishing we actually had devices that could translate animal sounds, you’re in for a treat. Researchers in Denmark have found a way to use AI to infer how pigs are feeling based on the sounds they make. 

Dr. Elodie Briefer, an expert in animal communication at the University of Copenhagen, co-led a team of scientists in the creation of an algorithm that can “decode” pig grunts and squeals. Briefer’s team began by collecting 7,414 high-frequency and low-frequency calls from 411 pigs across a variety of contexts, from birth to slaughter. 

The team then tested two automated methods of call classification: one a permuted discriminant function analysis (pDFA) designed by Roger Mundry, a biostatistician from Germany, the other a neural network capable of efficiently analyzing complex data. The neural network proved more effective at categorizing sounds than the pDFA due to its ability to classify calls according to the context in which each vocalization was produced. With this edge, the neural network was able to determine whether the pig calls were positive or negative with 91.5 percent accuracy. 

(Photo: Jeremy Bezanger/Unsplash)

As the researchers point out in their paper, which was published in the journal Scientific Reports earlier this week, previous attempts at decoding animal sounds have been extremely limited in scope, with the focus of individual projects being on “call types produced by animals of a given age, living in a specific environment and experiencing a limited number of well-defined situations.” This study, however, included pigs of every age, sex, body size, and “situation” and accounted for a variety of contexts: some pig calls were produced by pigs who were huddling together with littermates or approaching their mother (positive), while others were produced by pigs experiencing physical restraint or short-term social isolation (negative). 

Given the highly developed communication skills possessed by domestic pigs, it’s up to humans to find a way to understand what the average oink really means. Briefer’s team plans on continuing to develop their neural network to improve its accuracy; meanwhile, they’re hoping to develop a smartphone app that will help farmers decode pig calls and improve pigs’ quality of life.

Now Read:



sourse ExtremeTechExtremeTech https://ift.tt/X2CD0cK

ليست هناك تعليقات:

إرسال تعليق