From only one audio file, recognize speakers and separate each in different tracks