Existing methods for frequency analysis are generally based on applying
variations and extensions of Fourier mathematics to a given set of data
to determine the component frequencies in that set of data. Various
models have been proposed on how F0 could then be calculated but
limitations exist in being able to describe the psycho-physical results
reported by various researchers.
A different approach would use some common function for all frequencies
(hence wavelengths) at a given point in the stimulus to evaluate an
attribute or attributes for each frequency, and determine from complete
set of attributes the conditions necessary to identify F0. If such a
condition exists and was also capable of describing the psycho-physical
results, than one would have be encouraged in the ability of the
function to better explain the biology behind the auditory system.
To that extent, I have shown at my website, www.tonepitch.com in section
F0, that a function based on evaluative bivalence can provide the
conditions necessary to pick the correct F0, and also explain many of
the important psycho-physical phenomenon. The first of these is phase
locking of the evaluated attributes, with comments on its applicability
to explain dichotic hearing of pitch. I have also shown test results,
starting with virtual pitch using “resolved” and “unresolved” harmonics,
phase invariance with harmonic complexes, and finally results associated
with “mistuning” of a harmonic in a complex. Finally, I have given a
simple example of the problems associated with vowel recognition, but
that is just a flavor of the direction my research is taking, and maybe
somebody has a helpful suggestion for creation of a meta-model.
There are many psycho-physical results reported by the research
community, but I hope that what I have documented is of sufficient value
towards validating the approach that I have taken. Any comments would be
most helpful and I would be more than willing to update my website with
any other test results of general interest.
Thanking you for your patience in advance, cheers
Randy Randhawa
——————————
End of AUDITORY Digest – 7 Jan 2007 to 8 Jan 2007 (#2007-4)
***********************************************************
Amadeus does quite a good job of allowing flexibility in configuring its spectrograph (or sonogram as it¹s called) for specific sounds and it¹s quite affordable.
Some popular free software:
Praat a free software – is quite amazing for speech analysis purposes, both for the ease of its parameter control and for its automatic analysis functions (formants, pitch, intensity, pulses); however, it¹s not good for music for several reasons (1) it mixes stereo tracks into mono, (2) it does not offer logarithmic display, (3) it¹s stuck on the Gauss window which isn¹t flat (in fact it shows brown noise with equal intensity throughout the spectrum)
Audacity allows some level of control over window FFT size, color, and frequency range but uses linear display only, and worse of all, does not show the frequency you are pointing at with the cursor; it’s not really usable.
However, a free software that is quite impressive and available for both PC and MAC is Sonic Visualiser available at http://www.sonicvisualiser.org/index.html. It allows control over range, logarithmic/linear view, window type (7 options) and size, scale, color, threshold, gain and more.
Eldad