The PDF files below show four to six pitch periods from the center of 10 vowels taken from /hVd/ utterances produced by a male and female talker. The male and female speaker were randomly drawn from a 100-speaker database (50 males and 50 females) consisting of /hVd/ utterances recorded from speakers with a midwestern dialect typical of the Detroit, MI metropolitan area. The database resides in the Speech Perception Laboratory at the University of Pittsburgh-Johnstown run by Dr. John W. Mullennix. The stimuli were recorded, digitized, and edited using the TFR (Time-Frequency Representation) software from Avaaz Innovations, Inc.

The waveforms below were directly digitized and edited with Canadian Speech Research Environment (CSRE). The vowels were produced by a Male speaker from the Detroit metro area recorded randomly in an hVd format. These are 25 msec segments edited out from the central portion of these categorical vowel pairs. The graphics were furnished by Dr. John Mullennix. Dr. Mullennix's web site can be found at http://www.pitt.edu/~mullenni/.

There are several features that should be noted from the displays. First, notice the relatively smooth appearance of the vowel in each category that has a low F2 value (who'd and hood) compared to the rough, spiked appearance of the vowels with an high F2 value (heed and hid). Second, notice the number of F1 cycles per pitch period (each segment shows at least 3 pitch periods). Put simply, that is the number of low amplitude wave cycles between the higher amplitude, lower frequency waves. Remember that complex waveform displays are the result of several separate waves interacting together. Refer to Table I for specific formant values and the number of F1 cycles per pitch period (F1-F0/100).

Category 1
/i/ "heed"
/u/ "who'd"

Category 2
/I/ "hid"
/U/ "hood"