[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Music Transmission Index?
Michael and List Readers:
I don't know of an index for music that is analogous to the AI. A "music
transmission index," (presumably, this would predict the perceived fidelity
of music from acoustical measurements) would be useful in many applications,
particularly in my area of interest which is hearing aids. I would greatly
appreciate any references to review papers or standards describing metrics
used to quantify fidelity of equipment that transmits music because I am not
familiar with the fidelity literature, other than for speech and hearing
The articulation index rates the communication pathway on how well it
transmits speech sounds. An AI for music would indicate the quality of the
listening experience over a music transmission pathway, with the "golden
ear" listeners approving of AI = 1.0. Perceiving/recognizing/enjoying
music requires finer hearing ability than needed for recognizing speech;
even so, the AI calculation [Fletcher, H. and Galt, R.H., 1950, Journal of
the Acoustical Society of America 22: 89-151] may provide a useful starting
point and framework for devising a music index. This is because the AI
includes detection thresholds, loudness, critical bands, and other
psychoacoustic constants assembled in such a way so as to determine the
"amount" of speech "received," and a music index would need these constants
too (in some form). Then there would be more psychoacoustic data to account
for the greater demands on the ear required to process the music signal.
Maybe it wouldn't be so hard to construct an index for music. Experimental
work is required. Perhaps, the first step is to see how well the AI in its
present form can predict listeners' ratings of fidelity for conditions where
music is filtered or noise-masked. These are the conditions for which the
AI was designed. (Does anyone know of existing data like these?) We
already know that the "speech" AI is deficient for music because the AI is
limited to the frequency range 180-8000 Hz and music needs a wider range.
Even though there are a multitude of other difficult problems with applying
the AI to music-for example, how to quantify music to incorporated it into
the calculation (in the case of speech, the AI uses the long-term average
spectrum and level-distribution of a sample of connected speech)-I believe
constructing a music index is possible. The folks who developed the
original articulation index solved these problems for speech, and that seems
impossible to me even today.
Is anyone interested in or involved with music fidelity metrics? Please
----- Original Message -----
From: "Büchler Michael" <Michael.Buechler@xxxxxx>
Sent: Monday, January 03, 2005 4:09 AM
Subject: Music Transmission Index?
Recent postings about the AI, STI and SII made me wonder if there exists
something like a "Music Transmission Index".
Any comments and advice regarding previous work and references would be
Michael Buechler PhD
CH-8091 Zurich / Switzerland