README
This is a python project for the computer vision subject of computer science career of Universitat Autonoma de Barcelona
The voice identification project is made for a real problem, our group has been working in a robot that need a voice identification module. This module is not test in real conditions, is only a first model for our robot.
The notebook is a complete description of the proyect, in the module file we have put all the functions to preprocess and predict a audio file.
We have made some models that are explained in the notebook, all of there are saved in the modelos directory, if each one of you wants to train a model, you must load some audios and follow all the instructions of notebook.
The predict functions it could be correct to make a backtest of all the models made.
y_pred, clase_mayoritaria = predict( algorithm='windowing', dataType='ruido', model='rf', filter=True, audio='tu_archivo.mp3' )
There is more info in the notebook Other form to test an audio file is doing:
python PROYECTO/voiceident.py --algorithm specsModel --dataType ruidoNorm --model cnn --filter True --audio path/to/audio.mp3
All the combinations possibles there are explained in the notebook
Our best model we think is:
python PROYECTO/voideident.py --algorithm windowing --dataType ruido --model rf --filter True --audio path/to/audio.mp3
And his confusion matrix is: 
- Clone the repository:
git clone https://github.com/albertceballos0/UAB-VC-voice-identification.git - Navigate to the project directory:
cd PROYECTO - Install all the python dependencies:
pip install -r requeriments.txt
Contributions are welcome! If you find any issues or have suggestions for improvement, please open an issue or submit a pull request. We are gratefull to recive new featrues. Please submit a new pull request.
This project is made only for educational intention.