Animal-assisted therapy for persons with disabilities based on canine tail language interpretation via fuzzy emotional behavior model

Animal-Assisted Therapy (AAT) is the science that employs the merit of human-animal interaction to alleviate mental and physical problems of persons with disabilities. However, to achieve the goal of AAT for persons with severe disabilities (e.g. spinal cord injury and amyotrophic lateral sclerosis), real-time animal language interpretation is needed. Since canine behaviors can be visually distinguished from its tail, this paper proposes the automatic real-time interpretation of canine tail language for human-canine interaction in the case of persons with severe disabilities. Canine tail language is captured via two 3-axis accelerometers. Directions and frequency are selected as our features of interests. New fuzzy rules and center of gravity (COG)-based defuzzification method are proposed in order to interpret the features into three canine emotional behaviors, i.e., agitate, happy, and scare as well as its blended emotional behaviors. The emotional behavior model is performed in the simulated dog. The average recognition rate in real dog is 93.75% accuracy.