Project source can be downloaded from https://github.com/SBZed/speech-emotion-recognition.git
Saurabh Zinjad
All other known bugs and fixes can be sent to "zinjadsaurabh1997@gmail.com" with subject "speech emotion recognition Suggestion".
Reported bugs/fixes will be submitted to correction.
Our machine learning model tries to detect and predict various emotion in speech signal or human audio by detecting different features and component of speech affected by human emotion. Emotion detection from the speech is a relatively new field of research. Here, we are trying a new approach or ways to contribute towards emotion recognition research.
To get familiar and comfortable with project concept and outcomes please refer to this presentation.
If you want an explanation for the project watch this video.
- Go to this google drive link.
- Open speech_emotion_recognition.ipynb notebook. it's main code for project.
- For live demonstration or detecting your Speech emotion, Open live_demo_speech_emotion_recognition.ipynb notebook. Just keep two things in mind. First, You need to record your audio in very silent room without any noise.And secondly, Model is not perfect, Show the great result for Disgust, fear, Anger, Sad. But shows poor result for surprise, Happy.
To implement this project on your Machine
- Clone this repository.
- Go to terminal and type
pip install -r requirements.txt. - Then open ./Code/main_speech_emotion_recognition.ipynb and run it.
- Need to record audio in clear and silent environment.
- Dataset is not accurate and prominant. Even, human ear can't detect accurate emotion for some audios. Need Promising dataset.