Introduction - If you have any usage issues, please Google them yourself
This paper has presented a method based on non-negative matrix factorization for performing voice
activity detection that requires no training data the user and is robust to changes in the
noise environment.