[ Contents | Search | Next | Previous | Up ]
From: Klaas Bakker, Springfield MO, USA
Date: 29 Apr 2010
Time: 08:48:52 -0500
Remote Name: 188.8.131.52
Tiffany, Good point! The Talking Time measure if done by hand leads to good reliability if the sample is normal speech, and if there are no other judgments to be made (the traditional cluttering severity tool which has been online for five years now asks the clinician to track talking time, and cluttering time simultaneously!). The combined scoring has not been tested. All of this is reason for the new development to have talking time judged automatically through acoustic analysis. If such software has a good sample of background noise it is very capable (accurate, and reliable) in determining "talking time" by determining the time the speech signal is over a backround noise theshold level, and.... including silences less than 200ms (which could occur in running speech for example). I trust this methodology, but realize there might be a validity caveat in that people could be active in speech (planning, organizing, preparing) in which that take time and aren't necessarily producing sound! A subjective method might catch such moments and include them, and also leave out other moments when there is sound that has nothing to do with speech. In other words there are pros and cons. Having done the handscoring numerous times I know from experience that even if it is accurate and reliable... it is cumbersome! So my preference overall is the automated acoustic method giving the clinician full attention or identifying which portions of speech are perceived as cluttered... Hope this is the answer to you question?