With the growing interest in enhancing the well-being of pets through advanced technology, in this study, we address the challenge of accurately recognizing animal behavior using wearable sensors and machine learning techniques. Existing methods are often restricted to recognizing behavior in a single species, limiting their potential for general application across different animal types. This shortcoming hinders broader applications in multi-species environments and results in inconsistent monitoring outcomes. In this research, we present an improved approach for animal behavior recognition by developing a 1D Convolutional Neural Network Long Short-Term Memory hybrid model specifically designed to process inertial measurement unit sensor data. By targeting the most relevant movement features using accelerometer, gyroscope, and magnetometer data, our model achieves a high degree of precision in classifying common activities among cats and dogs, with recognition accuracies of 89% for cats and 94% for dogs. The results validate the applicability of our model in diverse contexts, making it a promising tool for enhancing automated behavior monitoring in animal–computer interaction. This research contributes to the development of intelligent systems that improve pet care and lay the foundation for broader applications in animal welfare and behavioral studies.