We Can Now Understand What Your Dog Feels, and We Have Filed the Patent for It

At Voicari, we have always believed that the bond between people and their animals deserves better tools. Today we are excited to announce an important milestone, the filing of our patent application for an AI-powered system that analyzes your dog's vocalizations, evaluates their physiological signals, and tells you in plain language how your dog is doing.

The problem we are solving

Dogs communicate constantly, through barking, whining, growling, and body language. Yet owners are largely left guessing: Is that bark excitement or fear? Is that whimper frustration? Existing solutions on the market rely on continuous audio recording, which drains the battery quickly and produces huge amounts of data without any real insight.

We wanted to build something fundamentally different: a system smart enough to listen only when it matters, and clever enough to understand what it hears.

"Understanding an animal's emotional state is essential for animal welfare, for addressing behavioral issues, and for strengthening the bond between animal and human."

What our patent application covers

Our patent-pending invention (application filed with the German Patent and Trademark Office (DPMA) on 27 February 2026, Voicari GmbH) describes a comprehensive method, device, and system for real-time monitoring and interpretation of an animal's physiological and acoustic signals. Here is what makes it unique:

🎙️ Smart listening. The microphone on the collar only activates when the sound level crosses a threshold. No continuous recording, no wasted battery.

🐾 AI right on the device. A lightweight machine learning model runs directly on the collar's processor and recognizes whether a sound is a bark before anything is transmitted to the cloud.

❤️ Multimodal sensing. Heart rate and movement data are combined with the acoustic analysis to build a complete picture of the emotional state.

🧠 Cloud AI pipeline. Our remote server runs a deep learning model with convolutional, recurrent, and attention layers for precise emotion recognition.

📱 Clear results. Emotional states and trend charts are displayed directly in our app, with AI-generated explanations and practical recommendations for pet owners.

🔒 Privacy by design. Only relevant bark events are ever transmitted. Your pet's data stays minimal, and your bandwidth stays untouched.

How it works: the technology explained simply

The Voicari wearable monitors the ambient sound level. As soon as your dog produces a sufficiently loud sound, the processor wakes up, records the audio, and immediately runs a local AI model: Is this a bark?

If it is, the audio data is securely transmitted to our cloud servers together with the physiological sensor data. There, a multi-stage AI pipeline takes over. First, the raw signal is converted into a log-mel spectrogram, a visual representation of the sound that highlights the frequency patterns crucial for animal vocalizations. For dogs, that is mainly the range below 4 kHz, where the emotional content is richest.

A first AI model filters out background noise and passes on only genuine animal vocalizations. A second, more powerful model, combining convolutional neural networks (CNNs) for spectral analysis with recurrent neural networks (RNNs) for temporal context, then classifies the emotion. An attention mechanism ensures the model focuses on the emotionally most telling parts of the bark.

The emotions we can recognize

Our system is trained to recognize 24 emotional states in dogs, organized into 7 core emotional families: Happiness/Play, Curiosity/Seeking, Care, Calm/Neutral, Fear, Anger/Rage, and Sadness/Panic/Grief.

These categories are based on established frameworks in affective science, adapted for animal vocalizations. The system also accounts for individual differences: your dog's age, breed, and size feed into the interpretation, which becomes more precise over time.

From the wearable to the app: the full journey

1. The wearable listens. As soon as a sound crosses the threshold, the device wakes up and starts recording.
2. On-device classification. A quantized ML model on the collar's processor decides: bark or not? Irrelevant sounds are discarded immediately.
3. Secure transmission. Confirmed bark events are securely transmitted to our cloud servers together with physiological sensor data.
4. AI pipeline processing. Our multi-stage AI pipeline preprocesses the audio file, extracts frequency-based features, filters vocalizations, and runs the emotion recognition.
5. Results in the app. The recognized emotional state is displayed in the Voicari app, with trend charts, daily summaries, and personalized recommendations.

Why this matters for animal welfare

Behind the technology lies a deeper motivation: the conviction that animals deserve to be understood. Stress and anxiety in animals often go unnoticed until they show up as behavioral problems or health complaints. With Voicari, pet owners get a continuous, non-invasive window into their animal's inner life and can respond to their needs faster and with more empathy.

Long-term tracking delivers real trends, not just snapshots. Is your dog more anxious on weekdays? Has their mood improved after a new routine? Over time, the app paints a complete picture, valuable information for owners and veterinarians alike.

What comes next

The patent application is a foundational step, but far from the end of the journey. We are continuously working to expand our training data, refine our emotion models, and extend support to more animal species. If you are a researcher, a veterinarian, or part of an animal welfare organization and interested in working together, we would love to hear from you.

And most importantly: stay tuned. The Voicari wearable is coming, and your dog has a lot to say.