Neural Networks Trained to Detect Emotion?

Are there any open-source Neural Networks that are trained to detect emotions?

I'd like to see if I can u开发者_高级运维se one to detect negativity from video/audio clips, and or text.

---

You could make one for audio or video, though I have to say I don't believe it would be particularly effective. Neural networks are great at pattern recognition. What pattern would a network tune itself to? What is the sound of 'happiness' or 'anger'?

---

This is the way I imagine it for clips - separating the problem into smaller ones:

1. You can focus on checking whether people are smiling in the video. First use an algorithm that detects faces. Then you create several random (in time) samples of these faces. Then you use another algorithm (trained neural network) to check if they are smiling in the form of a number: p of n samples smile detected.

2. Use voice recognition to check if they are laughing or crying.

3. Check nodding (beware in some countries it is the opposite way, but you can skip this part for now)

4. Check whether the clip starts loudly and ends that way.

Once you recognized these 4 (example ones) you can try to train a neural network on top of them:

smiling, positive nodding, loud start -> positive

A good sample data will be sport commentators announcing a team win/lost.

Actually all this is good when you do not know what are they saying and this is the case when you want to check all youtube videos, but you do not want to learn all the languages in the world.