ORCA: An AI-Driven Closed-Loop Animal Vocalization Classification and Signal Synthesis System

This paper presents the ORCA (Optimized Response to Creature Acoustics) device, a proof-of-concept AI pipeline that listens to animal sounds, classifies species and vocal type, maps vocalizations to emotional states, and synthesizes a meaningful acoustic response signal. Using the ESC-50 environmental sound dataset with targeted data augmentation, we train a two-layer Support Vector Machine classifier that achieves 78.1% overall species classification accuracy and a 0.96 F1-score on dog detection. We demonstrate a signal synthesizer that generates frequency-shaped acoustic responses corresponding to three dog vocal types: bark, growl, and whine. The complete system operates as a real-time closed loop: input audio → species detection → vocal classification → signal synthesis → playback. This work represents a first step toward a generalizable animal communication interface.