A comparison of cepstral coefficients and spectral moments in the classification of Romanian fricatives

Abstract

In this paper we explore two methods for the classification of fricatives. First, for the coding of the speech, we compared two sets of acoustic measures obtained from a corpus of Romanian fricatives: (a) spectral moments, and (b) cepstral coefficients. Second, we compared two methods of determining the regions of the segments from which the measures would be extracted. In the first method, the phonetic segments were divided into three regions of approximately equal duration. In the second method, Hidden Markov Models (HMMs) were used to divide each segment into three regions such that the variances of the measures within each region were minimized. The corpus we analyzed consists of 3,674 plain and palatalized word-final fricatives from four places of articulation, produced by 31 native speakers of Romanian (20 females). We used logistic regression to classify fricatives by place, voicing, palatalization status, and gender. We found that cepstral coefficients reliably outperformed spectral moments in all classification tasks, and that using regions determined by HMM yielded slightly higher correct classification rates than using regions of equal duration.

Acoustic classification of Russian plain and palatalized sibilant fricatives: Spectral vs. cepstral measures

“…We set out to compare two methods for fricative classification and, in doing so, to provide descriptive data regarding the 4-way contrast of Russian voiceless sibilant fricatives. Our findings were that cepstral coefficients outperformed spectral measures, regardless of the measure type (spectral moments alone, as in Spinu and Lilley (2016) cients versus spectral measures) and gender (MF -both genders collapsed, M only, F only).…”

“…Only the vocalic region adjacent to the fricative was used. The addition of vocalic data constitutes a notable difference compared to the previous analysis by Spinu and Lilley (2016).…”

“…In this study, we extend the method from Spinu and Lilley (2016) to a new language, Russian, focusing on a specific subset of fricatives -voiceless plain and palatalized sibilants. Adding to the complexity of the problem is the fact that this 4-way primary/secondary place contrast is extremely rare cross-linguistically.…”

Articulatory Characteristics of Secondary Palatalization in Romanian Fricatives

“…The stimuli analyzed for this study are a subset of a larger corpus collected to investigate several aspects of the Romanian inventory. Following the methodology in earlier studies [10,11,23] ] (cherry tree -cherry trees). The total for analysis was thus 3 fricatives × 4 words × 2 forms × 6 repetitions × 10 speakers = 1,440 tokens.…”

End-to-End Acoustic Feedback in Language Learning for Correcting Devoiced French Final-Fricatives

“…The fricative duration cue plays a significant perceptual role to distinguish between voiced and voiceless fricatives, and also lengthening the duration of the preceding vowel of the word-final fricative is an effective technique to produce a voiced fricative [16,10]. Low-frequency energy was included as a successful measure of voicing during the frication by a series of works [17,18]; here a spectrum over the whole frication noise was computed, and then the amplitude of the components below 500 Hz was measured.…”