Formant Frequencies Research Articles

Modeling vowel-inherent spectral change in varying consonant context

A wide variety of experimental evidence has shown that vowel inherent spectral change (VISC) is important in vowel identification. This evidence is drawn from production data, statistical pattern recognition, and perceptual experiments with both synthetic or manipulated naturally produced speech. Experimental studies often consider vowels in a constant consonant context which makes it difficult to factor out context effects from the VISC itself. In order to examine the magnitude of these consonant effects, we developed a statistical procedure inspired by Broad and Clermont [(2014). J. Phon. 47, 47–80] in which vowel formant frequencies are approximated by a linear combination of vowel and consonant influences which vary as a function of time. Vowels extracted from a database of both spontaneous and read speech were analyzed to produce context-normalized vowel-formant tracks. Results show that vowel formant frequencies vary systematically across their duration in both spontaneous and read speech and that all consonants in both onset and coda position show significant effects on vowel production across their entire duration. Although these formant patterns are seemingly complex, perceptual evidence suggests that listeners may only attend to onsets and offsets and that deviations from a straight-line interpolation between onset and offset must be relatively large for listeners to discriminate them.

ChildAugment: Data augmentation methods for zero-resource children's speaker verification.

The accuracy of modern automatic speaker verification (ASV) systems, when trained exclusively on adult data, drops substantially when applied to children's speech. The scarcity of children's speech corpora hinders fine-tuning ASV systems for children's speech. Hence, there is a timely need to explore more effective ways of reusing adults' speech data. One promising approach is to align vocal-tract parameters between adults and children through children-specific data augmentation, referred here to as ChildAugment. Specifically, we modify the formant frequencies and formant bandwidths of adult speech to emulate children's speech. The modified spectra are used to train emphasized channel attention, propagation, and aggregation in time-delay neural network recognizer for children. We compare ChildAugment against various state-of-the-art data augmentation techniques for children's ASV. We also extensively compare different scoring methods, including cosine scoring, probabilistic linear discriminant analysis (PLDA), and neural PLDA. We also propose a low-complexity weighted cosine score for extremely low-resource children ASV. Our findings on the CSLU kids corpus indicate that ChildAugment holds promise as a simple, acoustics-motivated approach, for improving state-of-the-art deep learning based ASV for children. We achieve up to 12.45% (boys) and 11.96% (girls) relative improvement over the baseline. For reproducibility, we provide the evaluation protocols and codes here.

Predicting the classification of high vowel sound by using artificial neural network: a study in forensic linguistics

<span lang="EN-US">One of the tasks in forensic linguistics, especially forensic phonetics, is evaluating the speech sounds in the recordings. The speech evaluation aims at identifying and verifying speakers to predict if the sound were spoken by the suspect or not. The common problem in the task is determining which acoustic features of the speech sounds are reliable for the speaker identification and verification. The purpose of this research is studying formant frequencies to predict high vowel sounds /i/, and /u/ by using artificial neural network (ANN). Using three various normalization methods (i.e., softmax, z-score and sigmoid), we utilized multilayer perceptron on backpropagation ANN with the architectural models of 4-5-2, 4-10-2 and 4-20-2. The results show that the z-score normalization method provides higher accuracy than the other two in all formations and the 4-10-2 formation has shown the highest accuracy (92.26%).</span>

Improved tactile speech perception using audio-to-tactile sensory substitution with formant frequency focusing

Haptic hearing aids, which provide speech information through tactile stimulation, could substantially improve outcomes for both cochlear implant users and for those unable to access cochlear implants. Recent advances in wide-band haptic actuator technology have made new audio-to-tactile conversion strategies viable for wearable devices. One such strategy filters the audio into eight frequency bands, which are evenly distributed across the speech frequency range. The amplitude envelopes from the eight bands modulate the amplitudes of eight low-frequency tones, which are delivered through vibration to a single site on the wrist. This tactile vocoder strategy effectively transfers some phonemic information, but vowels and obstruent consonants are poorly portrayed. In 20 participants with normal touch perception, we tested (1) whether focusing the audio filters of the tactile vocoder more densely around the first and second formant frequencies improved tactile vowel discrimination, and (2) whether focusing filters at mid-to-high frequencies improved obstruent consonant discrimination. The obstruent-focused approach was found to be ineffective. However, the formant-focused approach improved vowel discrimination by 8%, without changing overall consonant discrimination. The formant-focused tactile vocoder strategy, which can readily be implemented in real time on a compact device, could substantially improve speech perception for haptic hearing aid users.

Mantled howler monkey males assess their rivals through formant spacing of long-distance calls.

Formant frequency spacing of long-distance vocalizations is allometrically related to body size and could represent an honest signal of fighting potential. There is, however, only limited evidence that primates use formant spacing to assess the competitive potential of rivals during interactions with extragroup males, a risky context. We hypothesized that if formant spacing of long-distance calls is inversely related to the fighting potential of male mantled howler monkeys (Alouatta palliata), then males should: (1) be more likely and (2) faster to display vocal responses to calling rivals; (3) be more likely and (4) faster to approach calling rivals; and have higher fecal (5) glucocorticoid and (6) testosterone metabolite concentrations in response to rivals calling at intermediate and high formant spacing than to those with low formant spacing. We studied the behavioral responses of 11 adult males to playback experiments of long-distance calls from unknown individuals with low (i.e., emulating large individuals), intermediate, and high (i.e., small individuals) formant spacing (n = 36 experiments). We assayed fecal glucocorticoid and testosterone metabolite concentrations (n = 174). Playbacks always elicited vocal responses, but males responded quicker to intermediate than to low formant spacing playbacks. Low formant spacing calls were less likely to elicit approaches whereas high formant spacing calls resulted in quicker approaches. Males showed stronger hormonal responses to low than to both intermediate and high formant spacing calls. It is possible that males do not escalate conflicts with rivals with low formant spacing calls if these are perceived as large, and against whom winning probabilities should decrease and confrontation costs increase; but are willing to escalate conflicts with rivals of high formant spacing. Formant spacing may therefore be an important signal for rival assessment in this species.

Acoustic and Spectral Analysis of Adi Triphthongs

Adi, a zero-resource indigenous tribal language of Arunachal Pradesh, originated in Tibeto-Burman. According to the “UNESCO Atlas of the World's Languages in Danger of 2017” report, Adi is critically endangered. In this research, two new Adi triphthongs, /uai/ [uai] and /oai/ [ɔai], are proposed in the phoneme set of the Adi language. Adi has 51 phonemes, including 14 monophthongs (7 short and 7 long vowels), 19 diphthongs, two triphthongs, and 16 consonants. This research analysed Adi triphthongs’ spectral and acoustic properties, including intensity variation, long-term average spectrum (LTAS), and formant distribution. The mean of the first four formant frequencies of male and female Adi speakers is measured, which may be the identification of these two phones in the Adi speech recognition systems. Four different Adi words for each triphthong are recorded from 68 individual native Adi speakers (32 males and 36 females) in Arunachal Pradesh. This study may be helpful for phonetic modelling to build speech recognition applications for the Adi language.

Sentences used in the speech-to-song illusion: Comparisons of acoustic vowel space

In the speech-to-song illusion, certain spoken sentences start sounding like song when repeated several times. This perceptual transformation does not occur for all stimuli, suggesting that acoustic properties of the stimulus may contribute to the illusion. We investigated the contribution of the acoustic properties of vowels to this phenomenon by analyzing the acoustic vowel-space area bounded by formant frequencies of /i/, /æ/, and / ɔ/ in a dataset of transforming ( illusion) and non-transforming ( non-illusion) stimuli. In general, larger vowel-space areas are associated with more acoustic distinctions between vowel categories. We found that the overall vowel space was larger in non-illusion than illusion stimuli. A possible reason for this difference may be that listeners learn to associate large vowel spaces with speech and small vowel spaces with song through exposure to differences between formant frequencies in spoken and sung vowels. We propose that the shifted vowel spaces in which non-illusion sentences may be associated with speech perception thereby activate speech processing circuitry that inhibits the illusion of their transformation into song.

Auditory free classification of gender diverse speakersa).

Auditory attribution of speaker gender has historically been assumed to operate within a binary framework. The prevalence of gender diversity and its associated sociophonetic variability motivates an examination of how listeners perceptually represent these diverse voices. Utterances from 30 transgender (1 agender individual, 15 non-binary individuals, 7 transgender men, and 7 transgender women) and 30 cisgender (15 men and 15 women) speakers were used in an auditory free classification paradigm, in which cisgender listeners classified the speakers on perceived general similarity and gender identity. Multidimensional scaling of listeners' classifications revealed two-dimensional solutions as the best fit for general similarity classifications. The first dimension was interpreted as masculinity/femininity, where listeners organized speakers from high to low fundamental frequency and first formant frequency. The second was interpreted as gender prototypicality, where listeners separated speakers with fundamental frequency and first formant frequency at upper and lower extreme values from more intermediate values. Listeners' classifications for gender identity collapsed into a one-dimensional space interpreted as masculinity/femininity. Results suggest that listeners engage in fine-grained analysis of speaker gender that cannot be adequately captured by a gender dichotomy. Further, varying terminology used in instructions may bias listeners' gender judgements.

Acoustic and visual geometry descriptor for multi-modal emotion recognition fromvideos

<span>Recognizing human emotions simultaneously from multiple data modalities (e.g., face, and speech) has drawn significant research interest, and numerous research contributions have been investigated in the affective computing community. However, most methods concentrate less on facial alignment and keyframe selection for audio-visual input. Hence, this paper proposed a new audio-visual descriptor, mainly concentrating on describing the emotion through only a few frames. For this purpose, we proposed a new self-similarity distance matrix (SSDM), which computes the spatial, and temporal distances through landmark points on the facial image. The audio signal is described through an asset of composite features, including statistical features, spectral features, formant frequencies, and energies. A support vector machine (SVM) algorithm is employed to classify both models, and the final results are fused to predict the emotion. Surrey audio-visual expressed emotion (SAVEE) and Ryerson multimedia research lab (RML) datasets are utilized for experimental validation, and the proposed method has shown significant improvement from the state of art methods.</span>

Evaluating acoustic representations and normalization for rhoticity classification in children with speech sound disorders.

The effects of different acoustic representations and normalizations were compared for classifiers predicting perception of children's rhotic versus derhotic /ɹ/. Formant and Mel frequency cepstral coefficient (MFCC) representations for 350 speakers were z-standardized, either relative to values in the same utterance or age-and-sex data for typical /ɹ/. Statistical modeling indicated age-and-sex normalization significantly increased classifier performances. Clinically interpretable formants performed similarly to MFCCs and were endorsed for deep neural network engineering, achieving mean test-participant-specific F1-score = 0.81 after personalization and replication (σx = 0.10, med = 0.83, n = 48). Shapley additive explanations analysis indicated the third formant most influenced fully rhotic predictions.

The Impact of Trained Conditions on the Generalization of Learning Gains Following Voice Discrimination Training.

Auditory training can lead to notable enhancements in specific tasks, but whether these improvements generalize to untrained tasks like speech-in-noise (SIN) recognition remains uncertain. This study examined how training conditions affect generalization. Fifty-five young adults were divided into "Trained-in-Quiet" (n = 15), "Trained-in-Noise" (n = 20), and "Control" (n = 20) groups. Participants completed two sessions. The first session involved an assessment of SIN recognition and voice discrimination (VD) with word or sentence stimuli, employing combined fundamental frequency (F0) + formant frequencies voice cues. Subsequently, only the trained groups proceeded to an interleaved training phase, encompassing six VD blocks with sentence stimuli, utilizing either F0-only or formant-only cues. The second session replicated the interleaved training for the trained groups, followed by a second assessment conducted by all three groups, identical to the first session. Results showed significant improvements in the trained task regardless of training conditions. However, VD training with a single cue did not enhance VD with both cues beyond control group improvements, suggesting limited generalization. Notably, the Trained-in-Noise group exhibited the most significant SIN recognition improvements posttraining, implying generalization across tasks that share similar acoustic conditions. Overall, findings suggest training conditions impact generalization by influencing processing levels associated with the trained task. Training in noisy conditions may prompt higher auditory and/or cognitive processing than training in quiet, potentially extending skills to tasks involving challenging listening conditions, such as SIN recognition. These insights hold significant theoretical and clinical implications, potentially advancing the development of effective auditory training protocols.

Study on a Pig Vocalization Classification Method Based on Multi-Feature Fusion.

To improve the classification of pig vocalization using vocal signals and improve recognition accuracy, a pig vocalization classification method based on multi-feature fusion is proposed in this study. With the typical vocalization of pigs in large-scale breeding houses as the research object, short-time energy, frequency centroid, formant frequency and first-order difference, and Mel frequency cepstral coefficient and first-order difference were extracted as the fusion features. These fusion features were improved using principal component analysis. A pig vocalization classification model with a BP neural network optimized based on the genetic algorithm was constructed. The results showed that using the improved features to recognize pig grunting, squealing, and coughing, the average recognition accuracy was 93.2%; the recognition precisions were 87.9%, 98.1%, and 92.7%, respectively, with an average of 92.9%; and the recognition recalls were 92.0%, 99.1%, and 87.4%, respectively, with an average of 92.8%, which indicated that the proposed pig vocalization classification method had good recognition precision and recall, and could provide a reference for pig vocalization information feedback and automatic recognition.

Acoustic and Perceptual Effects of Articulation Exercises in Transgender Women.

This study measured the impact of articulation exercises using a cork and articulation exercises for lip spreading on the formant frequencies of vowels and listener perceptions of femininity in transgender women. Thirteen transgender women were recorded before and after the cork exercise and before and after the lip spreading exercise. Speech samples included continuous speech during reading and were analyzed using Praat software. Vowel formant frequencies (F1, F2, F3, F4, F5) and vowel space were determined. A listening experiment was organized using naïve cisgender women and cisgender men rating audio samples of continuous speech. Masculinity/femininity, vocal quality and age were rated, using a visual analogue scale (VAS). Concerning vowel formant frequencies, F2 /a/ and F5 /u/ significantly increased after the lip spreading exercise, as well as F3 /a/, F3 /u/ and F4 /a/ after the cork exercise. The lip spreading exercise had more impact on the F2 /a/ than the cork exercise. Vowel space did not change after the exercises. The fundamental frequency (fo) increased simultaneously during both exercises. Both articulation exercises were associated with significantly increased listener perceptions of femininity of the voice. Subtle changes in formant frequencies can be observed after performing articulation exercises, but not in every formant frequency or vowel. Cisgender listeners rated the speech of the transgender women more feminine after the exercises. Further research with a more extensive therapy program and listening experiment is needed to examine these preliminary findings.

Comparison of speech changes caused by four different orthodontic retainers: a crossover randomized clinical trial.

This study aimed to compare the influence of four different maxillary removable orthodontic retainers on speech. Eligibility criteria for sample selection were: 20-40-year subjects with acceptable occlusion, native speakers of Portuguese. The volunteers (n=21) were divided in four groups randomized with a 1:1:1:1 allocation ratio. The four groups used, in random order, the four types of retainers full-time for 21 days each, with a washout period of 7-days. The removable maxillary retainers were: conventional wraparound, wraparound with an anterior hole, U-shaped wraparound, and thermoplastic retainer. Three volunteers were excluded. The final sample comprised 18 subjects (11 male; 7 female) with mean age of 27.08 years (SD=4.65). The speech evaluation was performed in vocal excerpts recordings made before, immediately after, and 21 days after the installation of each retainer, with auditory-perceptual and acoustic analysis of formant frequencies F1 and F2 of the vowels. Repeated measures ANOVA and Friedman with Tukey tests were used for statistical comparison. Speech changes increased immediately after conventional wraparound and thermoplastic retainer installation, and reduced after 21 days, but not to normal levels. However, this increase was statistically significant only for the wraparound with anterior hole and the thermoplastic retainer. Formant frequencies of vowels were altered at initial time, and the changes remained in conventional, U-shaped and thermoplastic appliances after three weeks. The thermoplastic retainer was more harmful to the speech than wraparound appliances. The conventional and U-shaped retainers interfered less in speech. The three-week period was not sufficient for speech adaptation.

Назализованные гласные алтайского языка по данным назографии и спектрографии

The paper presents the examination of the vowel nasalization in the Altai language. The study involved an experiment on nasography. A native speaker of the Altai language recited a program comprising 180 lexemes in an isolated context, repeating it three times. The speech was captured by the icSpeech dualmicrophone system that allows sound from the mouth and nose to be recorded separately. The recordings were annotated using the Praat software and then analyzed with the Emu-SDMS corpus system using the R programming language. The nasality coefficient was calculated by comparing sound levels across channels. Vowels before nasal consonants exhibited a slow increase in nasality coefficient, gradually transitioning from a low to a medium level during vowel lengthening. Vowels following nasal consonants exhibited pronounced nasalization, with a heightened nasality level at the vowel onset and a consistent mid-level nasality throughout its duration. Vowels between two nasal consonants had a combination of these two types: high nasality level in the beginning, then nasality fall followed by slow growth. The average nasality coefficients for pure vowels were 6030 units, for prenasal vowels were 9860 units, for postnasal vowels were 16990 units, and for internasal vowels were 18290 units. Next, the spectra of nasalized and non-nasalized vowels were compared. The formant frequencies and levels were measured using the Praat software. The data shows a significant decrease in F1 level for nasalized vowels as opposed to non-nasalized vowels. The phenomenon can be attributed to the initial nasal antiresonance, correlating with acoustical models and typological data.

Characterization of Source-Filter Interactions in Vocal Vibrato Using a Neck-Surface Vibration Sensor: A Pilot Study.

Vocal vibrato is a singing technique that involves periodic modulation of fundamental frequency (fo) and intensity. The physiological sources of modulation within the speech mechanism and the interactions between the laryngeal source and vocal tract filter in vibrato are not fully understood. Therefore, the purpose of this study was to determine if differences in the rate and extent of fo and intensity modulation could be captured using simultaneously recorded signals from a neck-surface vibration sensor and a microphone, which represent features of the source before and after supraglottal vocal tract filtering. Nine classically-trained singers produced sustained vowels with vibrato while simultaneous signals were recorded using a vibration sensor and a microphone. Acoustical analyses were performed to measure the rate and extent of fo and intensity modulation for each trial. Paired-samples sign tests were used to analyze differences between the rate and extent of fo and intensity modulation in the vibration sensor and microphone signals. The rate and extent of fo modulation and the extent of intensity modulation were equivalent in the vibration sensor and microphone signals, but the rate of intensity modulation was significantly higher in the microphone signal than in the vibration sensor signal. Larger differences in the rate of intensity modulation were seen with vowels that typically have smaller differences between the first and second formant frequencies. This study demonstrated that the rate of intensity modulation at the source prior to supraglottal vocal tract filtering, as measured in neck-surface vibration sensor signals, was lower than the rate of intensity modulation after supraglottal vocal tract filtering, as measured in microphone signals. The difference in rate varied based on the vowel. These findings provide further support of the resonance-harmonics interaction in vocal vibrato. Further investigation is warranted to determine if differences in the physiological source(s) of vibrato account for inconsistent relationships between the extent of intensity modulation in neck-surface vibration sensor and microphone signals.

Formant Frequency Tuning of Three-Dimensional MRI-Based Vocal Tracts for the Finite Element Synthesis of Vowels

Formant Frequency Tuning of Three-Dimensional MRI-Based Vocal Tracts for the Finite Element Synthesis of Vowels

Production of English Vowel Contrasts in Spanish L1 Learners: A Longitudinal Study

The present study undertakes a longitudinal examination of forty postgraduate students, native Spanish speakers, during their first year at a UK university. The research focuses on both individual and collective progress in mastering distinctions within English vowel pairs (/iː/-/ɪ/, /ɪ/-/e/, and /uː/-/ʊ/), with a specific focus on adaptations towards achieving native-like English vowel pronunciation, particularly in terms of vowel quality. Prior research indicates that adult Spanish learners encounter difficulties in mastering the intricate linguistic nuances presented by English. The methodology involved recording Spanish-speaking participants reading a list of words (CVC context) at three different time points over a year. The analysis was based on formant frequencies using Praat, and Euclidean distances were calculated to represent the degree of separation between each pair of vowels. Information about external factors potentially influencing the development of vowel productions among speakers was gathered through a language background questionnaire. The outcomes suggested varying rates of advancement within the group, which could be attributed to the diverse levels of exposure and interaction with native English speakers during their year of study in the UK. These results affirm the learning processes in adult L2 production, emphasizing the critical role played by both the quantity and quality of time in the assimilation of pronunciations to novel L2 segments.

Beyond the average: embracing speaker individuality in the dynamic modeling of the acoustic-articulatory relationship

This paper explores the acoustic-articulatory relationship while considering individual differences in speech production. We aimed to determine whether there is a causal relationship between tongue movements and the contours of the first and second formant frequencies (F 1 and F 2) employing a hierarchical Bayesian continuous-time dynamic model, which allows for a more direct connection between the acoustic and articulatory measured variables and theories involving dynamicity. The results show predictive tendencies for both formants, where the anteroposterior and vertical tongue movements may predict changes in F 1, with rising predicting an increase and retraction a decrease; and with tongue fronting and tongue height inversely predicting F 2. Further, the modeled individual differences showed similar global tendencies, except for the rate of change of F 2. Overall, this study provides valuable insights into the relationship between tongue articulatory variables and formant contours, while accounting for between-speaker variability. The results show predictive tendencies for both formants, where the anteroposterior and vertical tongue movements may predict changes in f1, with rising predicting an increase and retraction a decrease; and with tongue fronting and tongue height inversely predicting f2. Further, the modeled individual differences showed similar global tendencies, however with larger error estimates. Overall, this study provides valuable insights into the relationship between tongue articulatory variables and formant contours, while accounting for between-speaker variability.

Using Formant Characteristics of Russian Vowels and Consonants for User Authentication

The article proposes a method for user authentication based on formant characteristics. A biometric image has been developed consisting of the formant characteristics of 46 sounds of the Russian language (6 vowels, 4 diphthongs and 36 consonants). An authentication system architecture based on formant characteristics and frequency of the leading formant of vowels and consonants is proposed. A Module for extracting the formant characteristics of vowels and consonants has been developed. A block diagram of the Module for extracting the formant characteristics of vowels and consonants and block diagrams of its submodules are presented. The Random Forest algorithm, k-nearest neighbors, logistic regression, multilayer perceptron and Support Vector Machine are combined into a cluster of algorithms. A block diagram of training and testing models based on a cluster of algorithms is given. Models were trained on the basis of algorithms clusters. The results of the experiment were obtained and their analysis was carried out.