Acoustic monitoring alerts farms to broiler welfare issues

An audio-based monitoring tool that uses bird vocalizations could serve as an early detection system for problematic respiratory disorders in poultry flocks.

Doughman Headshot3 Headshot
lilai | BigStock.com
lilai | BigStock.com

An audio-based monitoring tool that uses bird vocalizations could serve as an early detection system for problematic respiratory disorders in poultry flocks.

“Audio analytics non-invasively monitor a poultry flock for a wide variety of real time animal health and welfare metrics, during all phases of growth, at lower system cost than imaging-only. When used in conjunction with environmental sensors, this technology can directly validate animal welfare as a result of changes in environmental conditions (e.g. temperature, gas concentrations)," Tom Darbonne, CEO, AudioT, said.

Darbonne is one of the many experts scheduled to speak at the Virtual Poultry Tech Summit, scheduled for October 20-22, 2020. This one-of-a-kind online event facilitates the transition of innovation technologies from researchers and entrepreneurs into commercial applications for the benefit of the poultry industry. Make plans to attend and take a look at the future of the industry.

Registration for the 2020 Virtual Poultry Tech Summit is now open.

The challenges of respiratory diseases

Respiratory diseases are a serious animal welfare concern. Early detection offers more options for care and management, which can sometimes save the flock and reduce overall antibiotic usage.

“Automatic detection can put the veterinarian in the loop earlier – sometimes before the farmer – with better data about the actual condition,” Darbonne said.

The system uses a combination of acoustic signal processing techniques and machine learning to track and analyze the sound in the poultry houses. If an event of interest (e.g., coughs) or an anomaly (e.g, dry auger or fan motor failure) occurs, the farmers receive an alert.

“In general, developing effective machined learned models for non-speech audio is difficult, and requires a number of special techniques,” explained Darbonne. “One particular issue with poultry audio is called ‘the unbalanced data problem.’ To train a machine learning model takes quality data, and generally the more the better. In the instances of coughs and other vocalizations which are not common, training can skew towards the vocalizations which are more common.  We needed to develop techniques which work well with unbalanced data.“

The technology has already been tested in vaccine trials for laryngotracheitis and infectious bronchitis.

Like what you just read? Sign up now for free to receive the Poultry Future Newsletter.

Page 1 of 84
Next Page