Our emotions are an integral part of our lives, they usually reflect our opinions and our desires. It can be seen that there is a clear connection between emotions and human behavior. The ability to identify emotions provides many options in various aspects such as social media, market research, customer experience and more. As part of our work, we have fine-tuned a pre-trained model that trains on raw audio data and learns its contextual representation. With the help of fine-tuning, we can tweak this model to perform our task which is emotion recognition.
For better accuracy results, we decided to focus on 3 different class labels. Each class is a combination of common emotions.
- Positive - a mixture of Happy and Surprise.
- Neutral - a mixture of Neutral and Calm.
- Negative - a mixture of Anger, Fear, Sad, and Disgust.
- Clone this repository
- Enter the following command into your terminal:
python3 Main.py
- Python3
- Numpy
- Pytorch
OS: Ubuntu 20.04