EEC 201 Final Project
Created by Randall Fowler and Conor King
Speaker recognition refers to the classification of which person is speaking in an audio recording. This project uses Mel-Frequency Cepstrum Coefficients as features for a clustering algorithm. Each cluster will have a centroid to represent the center of the cluster, and each centroid is saved into a codebook.
For ease of use, many of the functions for this project have been implemented as class methods. The codelibrary will create, store, and reference codebooks created from training data. Runtime for the algorithm is relatively short as it may only take a minute or two to complete training or calculating accuracy of the model.
- numpy
- librosa
- matplotlib
- scipy
- tqdm
All audio data used for the project are stored within the Audio Files folder and codebooks will be stored in the Codebooks folder. All other necessary files can be found within the root director for the project.
To get a better understanding of the implementation of the project, you can take a look at the Preview.ipynb. This file was used as a starting point to create the processing method and start the tests.
To use the project, a good place to start would be Test 9 in the Tests.ipynb file. It shows how a codelibrary can be created. This is a collection of codebooks, and each codebook is created for individual training audio files in the specified folder.
If you wish to optimize on hyperparameters, there is a sweep.py that can be used to select different ranges of hyperparameters to search to find an optimal parameter set. The results of the sweep will be stored in the Codebooks folder as a pickle file. An example of viewing the result can be seen in the Preview.ipynb file.
Accuracy of different tests can be found in the Tests.ipynb file. To read about the project, there is a pdf file within the Paper folder. To watch a video relating to the project, you can find a mp4 file within the Video folder.
This project was completed for EEC 201 at UC Davis, Winter 2024, by Randall Fowler and Conor King.