An Automated Intervention to Reduce Shelter Dog Barking: A Hush Puppy Pilot Study

Abstract

Objective: Barking is a ubiquitous problem in animal shelters, with noise levels exceeding the Occupational Safety and Health Administration (OSHA)’s permissible exposure limits. This can cause significant health and welfare problem for workers, visitors, and the resident animals. The purpose of this study was to pilot and determine the efficacy of a computer-vision assisted classical conditioning procedure using automated food delivery with customizable automation scripts. Methods: Shelter dogs were selected for participation in the study from the Roanoke Regional Center for Animal Care and Protection and housed at the Virginia-Maryland College of Veterinary Medicine in three cohorts. An automatic feeder was installed on the kennels that was triggered by two separate detection systems: infrared beam breaks at the door to the ward and a computer vision system that detected human forms in the ward walkway. This system automatically delivered food contingent on the arrival or continued presence of a person in the ward. A sound-level meter was used to measure the noise in the ward, and from that the equivalent continuous sound level and percentage of time above 85 dB was calculated and monitored. Each cohort experienced at least three experimental conditions with the first two conditions implemented as non-concurrent, multiple baseline phases. Visual and statistical analysis were performed for each condition. Results: The method allowed for accurate and timely identification of people entering and lingering in the dogs’ ward and for contingent food delivery. This successfully eliminated issues with human compliance in implementing the intervention that were seen in previous studies. The largest noise level reduction was seen with novel, high-value food included in the feeders, resulting in a 3-dB reduction of the median noise level. The reduction in the median amount of time that barking produced noise levels above 85 dB ranged from 4% to 25%. Conclusions: The automation of a classical conditioning procedure using computer vision and automated food dispensers demonstrated that automated food delivery contingent on the arrival and continued presence of humans in the ward reduced the noise levels associated with barking. More work is required to reduce the noise levels below OHSA limits. Future studies can focus on optimizing the intervention parameters, the volume and type of food used, and integrating new technologies that may lead to more individualized and effective interventions.