Speech coding in the auditory nerve I: Vowel-like sounds. Delgutte, B and Kiang, N Y S The Journal of the Acoustical Society of America, 75(3):866-878.
Speech coding in the auditory nerve I: Vowel-like sounds [pdf]Paper  doi  abstract   bibtex   
Discharge patterns of auditory‐nerve fibers in anesthetized cats were recorded in response to a set of nine steady‐state, two‐formant vowels presented at 60 and 75 dB SPL. The largest components in the discrete Fourier transforms of period histograms were almost always harmonics of the vowel fundamental frequency that were close to one of the formant frequencies, the fundamental frequency or the fiber characteristic frequency (CF). For any fiber, the position of its CF relative to the formant frequencies (F1 and F2) appears to determine which of these components dominates the response. Specifically, the response characteristics of the tonotopically arranged array of fibers can be described in terms of five CF regions: (1) a low‐CF region below F1 in which the largest response components are the harmonics of the fundamental frequency closest to CF; (2) a region centered around CF = F1 in which the first formant and its harmonics are the largest components; (3) an intermediate region between F1 and F2 with prominent components at both the fiber CF and the fundamental frequency; (4) a region centered around CF = F2 in which harmonics close to the second formant are the largest for frequencies above the fundamental; and (5) a high‐CF region in which response spectra tend to show broad, multiple peaks at the formant and fundamental frequencies. These CF regions are related to the phonetic descriptions of vowels. For example, the extent of the low‐CF region is largest for ``open'' vowels (which have a high F1), and the intermediate region is distinct only for ``spread'' vowels for which F1 and F2 are more than 1.5--2 octaves apart. For all vowels, response activity for the majority of fibers is concentrated near the formant frequencies, in contrast to responses to broadband noise for which components near CF are dominant.
@article{delgutte_speech_1984,
	Author = {Delgutte, B and Kiang, N Y S},
	Date = {1984},
	Date-Modified = {2017-04-19 08:04:06 +0000},
	Doi = {10.1121/1.390596},
	Journal = {The Journal of the Acoustical Society of America},
	Keywords = {audiology, phonetics, speech perception},
	Number = {3},
	Pages = {866-878},
	Title = {Speech coding in the auditory nerve I: Vowel-like sounds},
	Url = {http://research.meei.harvard.edu/NeuralCoding/Papers/Delgutte84_I.pdf},
	Volume = {75},
	Abstract = {Discharge patterns of auditory‐nerve fibers in anesthetized cats were recorded in response to a set of nine steady‐state, two‐formant vowels presented at 60 and 75 dB SPL. The largest components in the discrete Fourier transforms of period histograms were almost always harmonics of the vowel fundamental frequency that were close to one of the formant frequencies, the fundamental frequency or the fiber characteristic frequency (CF). For any fiber, the position of its CF relative to the formant frequencies (F1 and F2) appears to determine which of these components dominates the response. Specifically, the response characteristics of the tonotopically arranged array of fibers can be described in terms of five CF regions: (1) a low‐CF region below F1 in which the largest response components are the harmonics of the fundamental frequency closest to CF; (2) a region centered around CF = F1 in which the first formant and its harmonics are the largest components; (3) an intermediate region between F1 and F2 with prominent components at both the fiber CF and the fundamental frequency; (4) a region centered around CF = F2 in which harmonics close to the second formant are the largest for frequencies above the fundamental; and (5) a high‐CF region in which response spectra tend to show broad, multiple peaks at the formant and fundamental frequencies. These CF regions are related to the phonetic descriptions of vowels. For example, the extent of the low‐CF region is largest for ``open'' vowels (which have a high F1), and the intermediate region is distinct only for ``spread'' vowels for which F1 and F2 are more than 1.5--2 octaves apart. For all vowels, response activity for the majority of fibers is concentrated near the formant frequencies, in contrast to responses to broadband noise for which components near CF are dominant.},
	Bdsk-Url-1 = {http://research.meei.harvard.edu/NeuralCoding/Papers/Delgutte84_I.pdf},
	Bdsk-Url-2 = {http://dx.doi.org/10.1121/1.390596}}
Downloads: 0