Text this: Endpoint detection enhancement for speaker dependent recognition