Laura Spinu

Palatalization in Romanian—Acoustic properties and perception

Abstract This paper presents the results of an acoustic study of fricatives from four places of articulation produced by 31 native speakers of Romanian, as well as those of a perceptual study using the stimuli from the acoustic experiment, allowing for a direct comparison between acoustic properties and perception. It was found that there are greater acoustic differences between plain and palatalized labials and dorsals as compared to coronals. The acoustic results were paralleled by the perceptual findings. This pattern departs from cross-linguistic generalizations made with respect to the properties of secondary palatalization. A likely source of the differences is the fact that previous studies of secondary palatalization typically involved stops which tend to exhibit various enhancement phenomena at the coronal place of articulation. Since the enhancement generally involves additional frication, this is not a useful strategy for fricatives at the coronal, or any other place of articulation. These findings form the basis of a discussion highlighting the differences between enhanced and non-enhanced secondary palatalization.

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 3674 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.

Vowel length in Luganda.

The relationship between vowel length and a following NC cluster in Luganda and other Bantu languages has received attention from both phonetic and phonological perspectives, however, the data are typically drawn from relatively small corpora. This study provides the results of the first relatively large-scale acoustic investigation of Luganda, using systematically constructed stimuli recorded by 10 native speakers of Luganda in Kampala, Uganda. Our results mostly confirm earlier findings and lend support to claims that the most appropriate analysis for Luganda requires a mismatch between the underlying representation of vowels and following NC clusters and their phonetic manifestation. Underlyingly, such strings consist of a short vowel followed by a moraic nasal coda, with the consonant in the onset of the next syllable (i.e. VN.CV). By contrast, the surface phonetic form contains a long vowel that is claimed to be lengthened in compensation for the nasals loss of moraic value, although the nature of t...

Classification of gender based on cepstral coefficients and spectral moments.

Several gender classification methods based on acoustic information were compared. The data came from 31 native speakers of Romanian (10 males, 21 females). A subset of fricatives and vowels (7348 tokens) was divided by hidden Markov model training into three acoustically uniform regions. For each region, (a) a set of cepstral coefficients, specifically c0–c4 and (b) two sets of spectral moments, specifically bark‐transformed and linear moments 1–4 plus rms, were extracted. The acoustic data were then used in a linear discriminant analysis to classify the tokens by gender. The findings show that cepstral coefficients perform better than spectral moments in gender classification, and that spectral moments are more successful than linear moments. The overall correct classification was 90% for the cepstral analysis, 78% for bark spectral moments, and 72% for linear moments. When the data from fricatives and vowels were examined separately, it was found that in all cases the vowel information yielded more acc...

Is there a bilingual advantage in phonetic and phonological acquisition? : The initial learning of word-final coronal stop realization in a novel accent of English

Research question: We address the question of whether the cognitive advantage of the bilingual mind, already demonstrated in the case of auditory processing or novel word acquisition, also applies to other linguistic domains, specifically to phonetic and phonological learning. Design: We compare the performance of 17 monolinguals and 25 bilinguals from Canada in a production experiment with two tasks: imitation and spontaneous reproduction of a novel foreign accent, specifically Sussex English. Data and analysis: To eliminate potential sources of variability, our focus is on a sound already existing in the subjects’ production (the glottal stop), but differently mapped to surface representations in the novel accent to which they were exposed (i.e., as an allophone of coronal stops in word-final position). We measured the glottal stop rates of our subjects in baseline, training, and post-training. Results: The two groups behaved differently, with bilinguals showing a larger increase of their glottal stop rate post-training. Our results are thus consistent with a bilingual advantage in phonetic and phonological learning. Originality: We interpret these findings in light of recent psycholinguistic work and conclude that echoic memory strategies, possibly underlain by stronger subcortical encoding of sound in bilinguals, may account for our results by facilitating the re-mapping between existing mental representations of sounds and existing articulatory command configurations. Significance: Our study adds to the body of work suggesting that there may be an advantage of bilingualism in second dialect learning in adulthood, and provides an explanation in terms of perceptual strategies in which echoic memory is involved. We also contribute to the body of research suggesting that imitation of an action can result in improved understanding of that action.

The Unit of Speech Encoding: The Case of Romanian.

The number of units in an utterance determines how much time speakers require to physically plan and begin their production [1]-[2]. Previous research proposed that the crucial units are prosodic i.e., Phonological Words (PWs), not syntactic or morphological [3]. Experiments on Dutch using a prepared speech paradigm claimed to support this view [4]-[5]; however, compounds did not conform to predictions and required the introduction of a different way of counting units. Since two PWs in compounds patterned with one PW, with or without clitics, rather than a phrase containing two PWs, a recursive PW’ was invoked. Similar results emerged using the same methodology with compounds in Italian [6], and it was thus proposed that the relevant unit for speech encoding is not the PW, but rather the Composite Group (CompG), a constituent of the Prosodic Hierarchy between the PW and Phonological Phrase that comprises both compounds and clitic constructions [7]. We further investigate the relevant unit for speech encoding using the same methodology in Romanian. Similar findings support the CompG as the speech planning unit since, again, compounds with two PWs pattern with single words and clitic constructions, not Phonological Phrases which also contain two PWs.

Automatic classification of English fricatives using cepstral coefficients

We use a classification tool previously tested on Romanian fricatives to categorize the front (non-sibilant) fricatives of English by place of articulation. Labiodental and interdental fricatives are difficult to distinguish acoustically, posing problems even for human perception. Prior classification work with English front fricatives has not been very successful with this contrast, with correct classification rates ranging from 40 to 60%. The feature set we use for coding the acoustic properties of the fricatives and their following vowels comprise the first six cepstral coefficients (c0–c5). The acoustic features are measured at 10-ms intervals across each segment; the measures obtained are then binned into three contiguous intervals for both the fricative and the vowel, representing the onset, steady state, and offset of each segment. The boundaries between regions are set by using a hidden Markov model to determine three internally uniform regions with respect to their acoustic properties. The mean v...

Identifying gender in fricatives: A comparison of human performance with a classification based on Cepstral coefficients.

A recent gender identification study using Cepstral coefficients extracted from voiced and voiceless fricatives from four different places of articulation yielded 78% correct classification in a linear discriminant analysis (LDA) [Spinu and Lilley (2010)]. This high accuracy was somewhat unexpected, as previous cross‐linguistic studies of fricatives generally found no significant effects of gender on various types of acoustic measures. The current study investigates whether humans can identify gender based solely on frication noise as successfully as the LDA. 30 English‐speaking undergraduates were asked to identify the gender of fricatives produced by 10 speakers (5 males and 5 females)—a subset of the sounds used in the LDA study. The analysis, currently in progress, suggests that humans may not be as successful as the LDA. These findings are expected to shed light on both (a) the differential distribution of acoustic features related to gender in fricatives, and (b) the extent to which these features a...

Classifying place of articulation and palatalization of fricatives based on cepstral coefficients and spectral moments.

Hidden Markov models (HMMs) with two types of acoustic features were used to classify fricatives by place of articulation and palatalization status. The data were recordings of 31 native speakers of Romanian who produced a total of 3674 fricatives. Segments from four places (labial, alveolar, postalveolar, and dorsal) were examined, each of which appeared as plain and palatalized (with a palatal secondary articulation). Both the fricatives and the preceding vowels were divided by HMM training into three acoustically uniform regions, corresponding to the three states of the HMM models. Separate sets of monophone HMMs were trained using (a) the first four spectral moments plus rms amplitude and (b) the first five Bark‐cepstral coefficients. Generally, the first and second regions/states of the fricatives were more important in classifying the segments by place, while the third state contributed more to the classification by palatalization status. The success of the classification depended on the specific co...

Palatalization in Romanian: Acoustics, perception, and the role of place of articulation.

Departing from the crosslinguistic generalization whereby the contrast between the plain and palatalized consonants is favored at the coronal, as compared to the labial, place of articulation, recent perceptual studies show native speakers of Romanian displaying higher sensitivity to this contrast in labials. To investigate this unexpected behavior, a production study was conducted with 31 subjects. Five plain and palatalized fricatives (/f, v, z, S, h/) were analyzed in terms of average duration and spectral properties (coefficients of the Bark Cepstrum). A linear discriminant analysis was run using duration and the Cepstral coefficients to predict segment type (plain/palatalized). 78.2% of 3674 tokens were classified correctly; however, the contribution of the duration and spectral properties of each segment showed interesting asymmetries. To summarize, /h/ was most successfully distinguished, /S/ was least successfully distinguished, and /f/ and /v/ were distinguished better than /z/. Our acoustic anal...