Skip to content

vggls/Speech_Emotion_Recognition_Classifier

Repository files navigation

SER_project_ML_Approach

This repo consists of :

  • SER_classification_task.ipynb Jupyter notebook which is a Speech Emotion Recognition (SER) study on the Crema-D dataset.

    The data can be obtained from the following Kaggle link : https://www.kaggle.com/ejlok1/cremad

    In order to view the plotly plots of the ipynb file you may use "Jupyter nbviewer" : https://nbviewer.org/

  • report.pdf which is a detailed report of the sections described in the ipynb file

  • presentation.pdf which is a short summary of the study's main conclusions

  • demo folder contains code to test new speech samples. The new sample must belong in one of the following classes : "happy", "neutral", "sad", "angry".

    The folder contains the following files :

    - demo.py : code to test new samples
    - model.pickle : contains the learned model and the audio features to choose from the test sample
    - scaler.pickle : contains the learned standard scaler
    - "test_sample" folder : folder to put the wav file to test
    

    In order to test a new sample you may insert the wav file in the "test_sample" folder and then run the "demo.py" file.

  • requirements.txt

  • pickles folder which contains all pickle files generated after running the ipynb file. They are included in the repo for saving purposes.

About

SVM classifier for speaker independent train-val-test sets

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published