Identification of high-pass filtered male, female, and child vowels: The use of high-frequency cues

Abstract

Vowels are characteristically described according to low-frequency resonance characteristics, which are presumed to provide the requisite information for identification. Classically, the study of vowel perception has focused on the lowest formant frequencies, typically F1, F2, and F3. Lehiste and Peterson [Phonetica 4, 161-177 (1959)] investigated identification accuracy of naturally produced male vowels composed of various amounts of low- and high-frequency content. Results showed near-chance identification performance for vowel segments containing only spectral information above 3.5 kHz. The authors concluded that high-frequency information was of minor importance for vowel identification. The current experiments report identification accuracy for high-pass filtered vowels produced by two male, two female, and two child talkers using both between- and within-subject designs. Identification performance was found to be significantly above chance for the majority of vowels even after high-pass filtering to remove spectral content below 3.0-3.5 kHz. Additionally, the filtered vowels having the highest fundamental frequency (child talkers) often had the highest identification accuracy scores. Linear discriminant function analysis mirrored perceptual performance when using spectral peak information between 3 and 12 kHz.

Gender identification from high-pass filtered vowel segments: The use of high-frequency energy

“…In other words, some listeners were more adept at identifying speaker gender from the high-pass filtered vowel segments than others, despite being matched for hearing sensitivity. This degree of variability was also observed in a previous study examining vowel category judgments from highpass filtered male, female, and child vowels (Donai & Paschall, 2015). The normal-hearing listeners in Donai and Paschall's study showed substantial differences in their abilities to identify vowel category from naturally produced, highpass filtered /hVd/ tokens from two adult male, two adult female, and two child speakers.…”

Automated Classification of Vowel Category and Speaker Type in the High-Frequency Spectrum

“…Vowels were recorded from two males, two females, and two children (one male and one female, both age 10) at 96 kHz and 24-bit resolution in an h/Vowel/d (hVd) context using a high fidelity microphone (Lawson 251) as described in Donai and Paschall. 3 Stimuli for the experiment consisted of five productions of six naturally produced, high-pass filtered hVd signals; /æ/ as in had, /i/ as in heed, /u/ as in who’d, /Ɔ/ as in hawd, //?// as in herd, and /e/ as in hayed . The hVd signals were recorded in connected speech using the carrier phrase, I say (hVd) again.…”

“…It was of interest to determine the effect, if any, sampling frequency had on classification accuracy, using a more traditional sampling frequency (48 kHz) and a less-commonly used, and higher, sampling frequency (96 kHz) used in Donai and Paschall. 3 For the second set, a 100 ms central portion of the vowel at a 48 kHz sample rate was used. To compute 13 dimensional MFCCs, window length and step time were set to 20 ms and 3 ms, respectively.…”

“…Similarly, it is assumed that identification of speaker gender from an audio signal relies heavily on the low-frequency region of the speech spectrum. 2 Recent evidence reported by Donai and Paschall 3 found that normal-hearing listeners were capable of identifying vowel category from six vowel tokens produced by two male, two female, and two child speakers with substantial accuracy (i.e., ranging from 40 to 70%) when low-frequency spectral peaks (at approximately 3.0-3.5 kHz) were removed via high-pass filtering. Classification results using linear discriminant function analyses from spectral peak information in the 3-12 kHz frequency region suggested that the signals were well-separated and inline with the perceptual data (in most cases within a few percentage points), thus supporting the perceptual findings.…”

“…The purpose of the following experiments is to determine the extent to which high-pass filtered vowel signals recorded in the Donai and Paschall 3 study could be separated using Mel frequency cepstral coefficients (MFCCs) in a supervised learning approach. MFCCs incorporate aspects of audition and are a popular method utilized in automatic speech recognition systems.…”