perm filename JH2[KI,ALS] blob
sn#097067 filedate 1974-04-14 generic text, type T, neo UTF8
Page 1
00100 The Stanford AI Pitch-Synchronous Fourier-Transform Formant Extractor
FOURIER
FORMANT
EXTRACTOR
Page 1
00300 The formant extractor is not a formant tracker in the usual sense since
FORMANT
EXTRACTOR
FORMANT
Page 1
00400 a fresh determination of the formant locations is made for each segment
FORMANT
Page 1
00600 rapid changes in formant location, particularly in the vicinity of
FORMANT
Page 1
00700 obstruants where the character of the obstruant is frequently revealed
OBSTRUANTS
OBSTRUANT
Page 1
00900 has been done is any attempt made to recogncile data for adjacent
RECOGNCILE
Corrected to: RECONCILE
Page 1
01200 Formant identification is based on the use of Fourier transforms using
FORMANT
FOURIER
Page 1
01400 zero crossing which preceeds the maximum excursion in amplitude.
PRECEEDS
Page 1
02000 cleanness and unwarrented broadening of the peaks in the spectrum because
UNWARRENTED
Page 1
02200 is a reasonable thing to do since the location of the formant peaks is
FORMANT
Page 1
02300 affected by the glottal loading during the latter part of the period
GLOTTAL
Page 1
02600 for his own pecular glottal loading effects since he attempts to produce
PECULAR
GLOTTAL
Page 1
02800 that the ear can do anything to diamiguate glottal coupling effects.
DIAMIGUATE
GLOTTAL
Page 1
02900 It is observed that this glottal loading effect is more pronounced
GLOTTAL
Page 1
03200 the closing of the glottis rather than lengthening the closed time
GLOTTIS
Page 1
03300 when they drop the pitch of their voice. A reasomable thing
REASOMABLE
Corrected to: REASONABLE
Page 1
03800 The location of the formant peaks
FORMANT
Page 1
04000 since windowing attenuates contributions to the transform from the
ATTENUATES
Page 1
04700 formants and the region below the usual lower limit for the first
FORMANTS
Page 1
04800 formant. These limits are shifted between male and female voices, but
FORMANT
Page 1
05200 that are of lessor amplitude. If the five points for the five formant
FORMANT
Page 1
05500 medial smoothing operation which will be discribed later.
MEDIAL
DISCRIBED
Page 1
05700 Since the ranges for the formants overlap, frequent conflicts occur
FORMANTS
Page 1
06200 Should the first and second formants identifications
FORMANTS
Page 1
06400 low frequence side extending the region to zero, and to the high
FREQUENCE
Corrected to: FREQUENCY
Page 1
06700 median values for the F1 and F2 regions are then compared. Actually
MEDIAN
Page 1
06800 a decision made on the basis of amplitude only, allowing a 6 db credit
DB
Page 1
07500 introduced by the resolution of the F1 F2 conflict or which maw have been
MAW
Page 1
08000 to be parobolic as determined from three data points these being that
PAROBOLIC
Page 1
08100 point at the maximum and points nearest the two three db down values.
DB
Page 1
09500 conflicts by the procedures just discribed. When this occurs the fai,lure
DISCRIBED
Page 1
09600 to locate a proper peak is signaled by storing a zero for the formant in
FORMANT
Page 1
09700 question and the program proceeds to the next formant. On the completion
FORMANT
Page 1
10000 formant in question by the value found for the previous time slot.
FORMANT
Page 1
10300 peaks are refined by parobolic interpolations based on the positions
PAROBOLIC
INTERPOLATIONS
Page 1
10600 needed, at least in the case of 512 point transforms on 20,000 hertz
HERTZ
Page 1
10800 the greatly improved smoothness of the resulting formant tracks seems
FORMANT
Page 1
10900 to indicate that a corresponding incease in accuracy has resulted.
INCEASE
Page 1
11100 The procedures so far discribed result in very good formant tracks.
DISCRIBED
FORMANT
Page 1
11600 due to more obscure reasons. In almost all cases these abnormalities
ABNORMALITIES
Page 1
11800 a final process of medial smoothing. This is done in one direction only,
MEDIAL
Page 1
11900 going forward in time each value for each formant is replaced by the
FORMANT
Page 1
12000 median value of the point in question, its predisesor (as already
MEDIAN
PREDISESOR
Page 1
12400 the effect of correcting true extrema but an extrema which persists for
EXTREMA
EXTREMA
Page 1
12500 but a single pitch period probably does not contain much phonetic
PHONETIC
Page 1
12700 true extrema by applying the medial smoothing only to points that
EXTREMA
MEDIAL
Page 1
12800 lie more than, say, 2 db away from their nearest neighbor. This
DB
Page 1
13100 The advantages of this method of formant extraction over other more
FORMANT
EXTRACTION
Page 1
13300 results in the vicinity of obstruents where the rapid changes in formant
OBSTRUENTS
FORMANT
Page 1
13500 of the obstruent is contained in this transition region.
OBSTRUENT