cagbal/bow-speech-keyword-detector

A new keyword detector for MATLAB using Bag of Words Method

A Keyword Detection Algorithm using Bag of Words

What is this repository for?

Summary The project is a MATLAB code for detecting a keyword in a short speech and is done for one my master courses in Bogazici University. Therefore, the code is really new and the tests are not completed. Algorithm works well if it is tested with voices of same people in the training database. To make it more general, you can add more and more speech data to training set and see the effect of it.

• Version 1.0

Getting Started

• You need to download the mfcc database(link provided below) and a database to get started. Then, follow the instructions in code such as how to set your dataset and show it to code. After, run the code by using run.m file.

Contribution guidelines

• All the contributions are welcome, because the code needs to be improved so much.

Dataset

• I used the CMU_ARCTIC_DATABASE, but you can use which database you want, but the SNR ratio must be high in your dataset.

Credits

• The code is using the mfcc toolbox written by Kamil Wojcicki and can be downloaded from link below:
• http://www.mathworks.com/matlabcentral/fileexchange/32849-htk-mfcc-matlab

LICENSE

• MIT