This study addresses a key limitation in animal-assisted therapy (AAT) for individuals with severe disabilities, including spinal cord injury, stroke, and amyotrophic lateral sclerosis. While AAT can significantly improve mental and physical well-being, effective participation requires a mechanism to interpret animal behavior in real time, especially when verbal or physical interaction is limited.
The authors focused on canine tail language, a highly expressive and externally observable signal of emotional state. Tail movements were recorded using two 3-axis accelerometers, allowing continuous capture of movement direction and oscillation frequency. These parameters served as the primary features for emotional state classification.
To translate sensor data into meaningful emotional categories, the study proposed a novel fuzzy inference system based on Gaussian–Trapezoidal membership functions combined with a center-of-gravity (COG) defuzzification method. This framework mapped tail movement features into four core canine emotional behaviors: agitate, happy, scare, and neutral, while also accounting for blended emotional states that naturally occur in real interactions.
The emotional behavior model was first validated using a simulated dog and subsequently evaluated with a real dog. In both contexts, the system achieved a perfect recognition rate, demonstrating high robustness and practical feasibility for real-world deployment.
By enabling automatic, real-time interpretation of canine emotional signals, this approach enhances human–canine communication in therapeutic settings and supports more inclusive, adaptive, and ethically grounded animal-assisted therapy. The study illustrates how wearable sensing and fuzzy logic can bridge communication gaps, improving welfare outcomes for both humans and dogs.
Source: human-N Animal-Assisted Therapy for Persons with Disabilities Based on Canine Tail Language Interpretation via Gaussian-Trapezoidal Fuzzy Emotional Behavior Model. Medicine & Engineering. (Publication details as provided).
