Related Experiment Videos
Multi-QuadEmoNet: cat and dog emotion classification model from animal vocalization using multi-stage LSTM-GRU
Sundara Sobitha Raj Anubha Pearline1, Geetha Subbiah1, Ashley Sanu1
1School of Computer Science and Engineering, Vellore Institute of Technology Chennai Campus, Chennai, Tamil Nadu, India.
Frontiers in Veterinary Science
|July 8, 2026
Summary
This study introduces a novel Multi-Quadro-Emotion Network (Multi-QuadEmoNet) for recognizing animal emotions from vocalizations. The system achieves high accuracy in distinguishing and classifying cat and dog emotions, enhancing human-pet relationships.
Area of Science:
- Artificial Intelligence
- Animal Behavior
- Machine Learning
Background:
- Recognizing pet emotions from vocalizations is challenging with traditional methods.
- Understanding pet emotions is crucial for a positive human-animal relationship.
- Existing methods lack the sophistication to decode complex animal vocal cues.
Purpose of the Study:
- To develop an advanced animal emotion voice recognition system.
- To utilize a multi-modal approach for enhanced accuracy.
- To improve human-pet companionship through emotional understanding.
Main Methods:
- A novel Multi-Quadro-Emotion Network (Multi-QuadEmoNet) was proposed.
- Mel-Frequency Cepstral Coefficients (MFCCs) were used for audio feature extraction.
- Deep learning models (QuadEmoNet) were trained on a custom dataset of 8,000 cat and dog audio samples.
Main Results:
- The Multi-QuadEmoNet achieved a highest accuracy of 95% (DMAD) and 90% (CMAD).
- The system successfully classified general cat/dog vocalizations and specific emotions for each species.
- A user interface was developed for real-time animal emotion prediction.
Conclusions:
- The developed system demonstrates high efficacy in animal emotion recognition.
- This technology can significantly improve human-animal interaction and welfare.
- The Multi-QuadEmoNet offers a promising direction for affective computing in animals.
