[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
I would like to see clearer about the intelligibility of sine-wave speech.
The original Science study of Remez et al. uses 18 test subjects and they
report on the average 2.8 syllables transcribed correct (of a sentence of
7 syllables). I tried to run a transcription experiment similar to theirs,
but I got much-much worse results (practically, I could hardly find 1-2
persons from about 10 who could comprehend one word). Then, in the PhD
thesis of Jon Barker I found that "there would appear to exist a
considerable proportion of the population who are unable to hear SWS as
speech at all. In most SWS studies a pretest is employed to remove these
subjects from the experiment. The number of subjects removed can be as
high as 40%".
First of all, I would like to know if anyone has similar results but with
many more subjects (as only increasing the number of samples could help
reduce the variance of the measured transcription rates). it would be good
to see if others could replicate these preformance scores or got worse
Secondly, if my results are worse then they should be, I wonder why.
Is it because
- I was unlucky with the choice of subjects (however, Remez et al. did not
mention "pre-filtering" the subjects).
- The choice of the sentence was unlucky (the sentence they used was
deliberately chosen to contain almost only wovels and glides).
Hungarian Academy of Sciences *
Research Group on Artificial Intelligence * "Failure only begins
e-mail: firstname.lastname@example.org * when you stop trying"