Thanks to machine learning techniques, audio software can help farmers detect sick birds before they can.

As part of the Virtual Poultry Tech Summit, Tom Darbonne, CEO of AudioT, and Brandon Carroll, the co-founder of AudioT, spoke on October 22, 2020, about the advances made by their technology as well as its potential future uses.

In March 2020, AudioT received a $2 million grant as one of the winners of the SMART Broiler research initiative. The program, supported by the Foundation for Food and Agriculture Research and McDonald’s Corp., is designed to aid the development of automated monitoring tools to assess broiler welfare.

Origins of the technology

Darbonne said the company began in 2019 but the concept of detecting poultry diseases with audio is at least a decade old. AudioT wanted to start with a focus on laryngotracheitis and infectious bronchitis because outbreaks of the infections can have catastrophic implications for poultry farmers and vaccinations are not a complete solution for the disease.

These two diseases are primarily detected by a farmer present in the house who hears the coughing of the birds. The amount of days that pass after an outbreak are critical in the progression of the disease and its potential damage to a flock. If there were a way to automatically detect it instead of waiting for a farmer to randomly encounter it, it could make detection of the disease – and treatment – quicker.  

How it works

Therefore, AudioT set out to teach a machine how to detect the disease and effectively provide constant surveillance. Darbonne said the challenge was to create a hardware and software system to detect and recognize the specific sounds of an avian cough symptomatic of the conditions among the noise of the poultry house.

The research was able to identify certain auditory shapes in the sonic spectrum of the house that correlated with the coughs. The next challenge was to teach a machine to recognize those shapes and share the detection with its user.

Advertisement

Carroll explained that using machine learning, or teaching computers to recognize and respond to patterns, it was able to build software that could sort out the signals – coughing – from the noise – all the other normal sounds of the chicken house.

Future applications

Darbonne said the underlying technology possesses great potential for future uses for both animal health and bird welfare. As part of the SMART Broiler challenge, he said, designers are asked to develop devices that minimize cost, need for processing capacity and need for high speed internet access.

Audio surveillance is a great contender because it meets many of those requirements. Microphones are well established technology, durable for live production environments, high compatibility with other sensory systems and can be added relatively cheaply. The audio processing software requires much less processing capacity than video or other imaging and therefore could be more widely used.

Now that the machine learning patterns are established, learning to recognize the other sounds chickens make – whether they are associated with play, pain, fear, fatigue, pleasure, hunger, etc. – would be relatively easy. This could open the door for a myriad of additional animal health and bird welfare applications.

Darbonne said the audio technology can detect other factors like, food shortages, sleep cycles, instances when water pressure in the lines are too low, when ventilator fans are about to fail and other unseen or otherwise undetected issues. When combined with other environmental sensors, like thermometers and air quality detection devices, the machinery can detect a range of other issues, too.

Attend the 2021 Poultry Tech Summit

Join an exclusive international gathering of industry-changing innovators, researchers, entrepreneurs, technology experts, investors and leading poultry producers at the 2021 edition of Poultry Tech Summit on October 31 - November 2 in Atlanta, Georgia. Attendees can expect the same groundbreaking innovation and insightful presentations that made the previous events well-attended with deep dialogue on new prospective solutions and developing technologies.