Forensic phonetics

Предмет докторске дисертације јесте анализа вокалских јединица у енглеско-српском међујезику говорника чији се матерњи дијалекат разликује. Посебан фокус је на опсервацији монофтонга општеамеричког варијетета, одн. вокала /ɪ, i, ʊ, u, ɛ, æ, ʌ, ɑ, ɔ/ у међујезику испитаника пореклом из косовско-ресавске, шумадијско-војвођанске и херцеговачко-крајишке дијалекатске зоне. Три групе говорника чији је говор анализиран формиране су на основу дијалекатског порекла учесника, док су у погледу осталих варијабли групе хомогене. Задаци за прикупљање корпуса подразумевају читање речи које илуструју посматране вокале, а које су уметнуте у оквирне реченице на српском и енглеском језику. Истраживање се темељи на доминантним теоријским моделима о усвајању фонолошких система, тј. Ревидираном моделу учења говора (Флеги, Бон 2021) и Моделу перцептивне асимилације – Ј2 (Бест, Тајлер 2007). Примарни циљ дисертације је испитивање интерференције Ј1 дијалекта на усвајање Ј2 вокалских категорија.
Резултати анализе показали су да постоји међудијалекатска неуједначеност у продукцији одређених матерњих гласова. Конкретно, код говорника косовско-ресавског и шумадијско-војвођанског дијалекта та разлика тиче се отворенијег изговора кратких вокала /о/ и /u/ код косовско-ресавских испитаника, док изговор говорника из косовско-ресавске и херцеговачко-крајишке зоне разликује отвореност вокала /е/, /о/ и /uː/. Када је реч о положају Ј2 јединица у вокалском простору, већа варијабилност примећује се унутар сваког дијалекта него између њих, што указује на слаб дијалекатски утицај. Испитаници углавном теже да успоставе квалитативну разлику између Ј2 и Ј1 категорија, иако квалитет Ј2 вокала није сасвим адекватан. Амерички вокали /ʌ/, /ɛ/ и /i/ су код великог броја говорника супституисани српским вокалима /а/, /е/ и /iː/. За остале вокале може се констатовати да је дошло до формирања композитног квалитета, будући да су F1 и F2 вредности између типичних вредности за Ј1 и Ј2 вокале. Најчешћи типови асимилације јесу некатегоризовани–категоризовани тип (енгл. UC type), комбинација два некатегоризована гласа (енгл. UU type) и асимилација два Ј2 гласа у исту Ј1 категорију (енгл. CG type). Неадекватан вокалски квалитет, као и уочене супституције Ј2 вокала вокалима из Ј1 показују да у међујезику наших говорника постоји узајаман Ј1–Ј2 утицај.

aparezcan nuevas necesidades relacionadas con el acceso de un número cada vez mayor de personas a sistemas informáticos que tradicionalmente habían estado sólo al alcance de una minoría. Operaciones tan cotidianas como comprar un billete de tren o de avión en una máquina expendedora o retirar dinero de un cajero requieren un cierto grado de "alfabetización" informática, y prueba de ello es que no son pocos los usuarios que experimentan dificultades cuando se enfrentan por primera vez a un sistema automático. Existen también personas que, por su edad, su formación o por tener alguna discapacidad, no pueden utilizar un teclado y una pantalla, hasta ahora la forma más habitual de interacción con un ordenador. La extensión del uso del teléfono a capas cada vez más amplias de la población ha favorecido la creación de servicios telefónicos para atender las demandas de los consumidores. Así, han surgido la banca telefónica, los portales de voz desde los que se puede obtener información similar a la que aparece en los portales en la web, los centros de atención al cliente que deben funcionar ininterrumpidamente, los servicios de reserva de entradas para espectáculos, etc. En muchos de estos casos no es factible -o, cuanto menos, no es económico-contar con un operador humano, por lo que es preciso diseñar instrumentos que proporcionen de un modo automático los servicios requeridos. Además, el pequeño tamaño de aparatos como los teléfonos móviles y de los llamados asistentes personales (PDA) no permite incorporar un teclado y una pantalla en la que se pueda leer o escribir cómodamente un texto, y tampoco es práctico obligar al usuario a llevar consigo -pese a que existen en el mercado-una serie de accesorios para obtener con facilidad la información deseada. Una posible solución a estos problemas parece venir de la mano de las nuevas interfaces multimodales, como las pantallas táctiles, el reconocimiento de la escritura manuscrita o, muy especialmente, el reconocimiento del habla. No deja de resultar hasta cierto punto paradójico que, siendo el habla el medio de comunicación por excelencia entre las personas, la interacción con los equipos informáticos se lleve a cabo casi siempre mediante la escritura. Las denominadas tecnologías del habla tienen como finalidad facilitar el uso de los ordenadores introduciendo y recibiendo información de modo oral,

LLISTERRI, J.- CARBÓ, C.- MACHUCA, MJ-de la MOTA, C.- RIERA, M.- RÍOS, A. (2003) "El papel de la lingüística en el desarrollo de las tecnologías del habla", in CASAS GÓMEZ, M. (Dir.) - VARO VARO, C. (Ed.) VII Jornadas de Lingüística. Cádiz: Servicio de ...

Resumen: VILE (Estudio acústico de la variación inter e intralocutor en español) es un proyecto de investigación básica, financiado por el Ministerio de Ciencia y Tecnología (BFF2001-2551, 2001-2004), cuya finalidad es el estudio acústico de la variación fonética ...

This article aims to describe the phenomena that characterize Poles as speakers of peninsular Spanish (specifically in
its Granada’s variation) in their pronunciation of vowels. The analysis bases on surveys collected in the non-bilingual adult Polish
population residing in Granada for more than 10 years. The investigation includes the representation and the phonetic and acoustic
comparison of the vowels in their most complexes contexts. The study describes the different ways of vowel performance in Spanish by
Polish immigrants. The conclusions reveal the degree of its phonetic assimilation in specific cases, as well as the parameters influencing
its null adjustment, partially or totally akin to the native vowel pronunciation

The most frequent way in which criminals disguise their voices implies changes in phonation types, but it is difficult to maintain them for a long time. This mechanism severely hampers identification. Currently, the CIVIL corpus comprises 60 Spanish speakers. Each subject performs three tasks: spontaneous conversation, carrier sentences and reading, using modal, falsetto and creak(y) phonation. Two different recording sessions, one month apart, were conducted for each speaker, who was recorded with microphone, telephone and electroglottography. This is the first (open-access) corpus of disguised voices in Spanish. Its main purpose is finding biometric traces that remain in voice despite disguise.

Defining the relevant population to sample is an important issue in data-based implementation of the likelihood-ratio framework for forensic voice comparison. We present a logical argument that because an investigator or prosecutor only submits suspect and offender recordings for forensic analysis if they sound sufficiently similar to each other, the appropriate defense hypothesis for the forensic scientist to adopt will usually be that the suspect is not the speaker on the offender recording but is a member of a population of speakers who sound sufficiently similar that an investigator or prosecutor would submit recordings of these speakers for forensic analysis. We propose a procedure for selecting background, development, and test databases using a panel of human listeners, and empirically test an automatic procedure inspired by the above. Although the automatic procedure is not entirely consistent with the logical arguments and human-listener procedure, it serves as a proof of concept for the importance of database selection. A forensicvoice-comparison system using the automatic database-selection procedure outperformed systems with random database selection.

Soft and far-field voices-often overlooked or dismissed as background noise can be effectively recorded with most modern audio-capturing devices. Recent advances in artificial intelligence (AI), particularly in voice enhancement, psychoacoustic modeling, and fingerprint analysis, have made it feasible to identify individuals through their vocal idiosyncrasies with precision. This paper explores how these tools can serve forensic and legal purposes, including the detection of aggravated harassment and behaviors masked under terms like 'telepathy'. The analysis shows that individuals claiming or simulating exceptional cognitive or communicative abilities (e.g., alleged mentalists or telepaths) can be identified and held accountable under existing laws. Such methods dramatically reduce the cost compared to traditional forensic audio analysis.

LLISTERRI, J.- CARBÓ, C.- MACHUCA, MJ-de la MOTA, C.- RIERA, M.- RÍOS, A. (2003) "El papel de la lingüística en el desarrollo de las tecnologías del habla", in CASAS GÓMEZ, M. (Dir.) - VARO VARO, C. (Ed.) VII Jornadas de Lingüística. Cádiz: Servicio de ...

A speech database has been collected for use to highlight the importance of "speaker factor" in forensic voice comparison. FABIOLE has been created during the FABIOLE project funded by the French Research Agency (ANR) from 2013 to 2016. This corpus consists in more than 3 thousands excerpts spoken by 130 French native male speakers. The speakers are divided into two categories: 30 target speakers who everyone has 100 excerpts and 100 "impostors" who everyone has only one excerpt. The data were collected from 10 different French radio and television shows where each utterance turns with a minimum duration of 30s and has a good speech quality. The data set is mainly used for investigating speaker factor in forensic voice comparison and interpreting some unsolved issue such as the relationship between speaker characteristics and system behavior. In this paper, we present FABIOLE database. Then, preliminary experiments are performed to evaluate the effect of the "speaker factor" and the show on a voice comparison system behavior.

A speech database has been collected for use to highlight the importance of “speaker factor” in forensic voice comparison. FABIOLE has been created during the FABIOLE project funded by the French Research Agency (ANR) from 2013 to 2016. This corpus consists in more than 3 thousands excerpts spoken by 130 French native male speakers. The speakers are divided into two categories: 30 target speakers who everyone has 100 excerpts and 100 “impostors” who everyone has only one excerpt. The data were collected from 10 different French radio and television shows where each utterance turns with a minimum duration of 30s and has a good speech quality. The data set is mainly used for investigating speaker factor in forensic voice comparison and interpreting some unsolved issue such as the relationship between speaker characteristics and system behavior. In this paper, we present FABIOLE database. Then, preliminary experiments are performed to evaluate the effect of the “speaker factor” and the...

The present work aims to provide more data regarding the comparison and discrimination of speech samples in the field of forensic phonetics. The methodology used was an acoustic method called long term average spectrum (LTAS), by calculating its spectral decline. The informants were grouped into a total of four female speakers of the same age by close geographical origins. The extracted data were presented qualitatively and quantitatively. In view of the results, the identification of speakers was not
successful in all cases. After discussing the results, implications for judicial phonetics and recommendations for future studies were collected.

Aunque los fenómenos de hesitación en el discurso --pausas sonoras y alargamientos-- se han abordado desde diferentes puntos de vista, son escasos los estudios que analizan sus manifestaciones acústicas en español. En este trabajo se presenta una caracterización preliminar de las hesitaciones que, generalmente, suelen aparecer en el habla espontánea con diferentes funciones discursivas. Se han analizado 180 minutos de conversaciones en las que han participado seis hablantes, con una duración aproximada de 30 minutos cada una, y 30 fragmentos de producción espontánea de un minuto de duración correspondientes a tres sesiones de grabación de 10 locutores. En las primeras se da una interacción entre los participantes, mientras que en las segundas, el hablante sabe que no será interrumpido durante su intervención. El análisis acústico se ha complementado con un estudio perceptivo en el que los oyentes debían valorar la naturalidad y la espontaneidad de los mismos fragmentos presentados con o sin alargamiento y con o sin pausa sonora. Los resultados muestran, en primer lugar, que la pausa sonora más frecuente en español es [e?]. Los valores medios y las desviaciones correspondientes a la duración de los alargamientos y de las pausas sonoras no arrojan diferencias significativas entre estas hesitaciones. No obstante, el rango de los valores medios debería considerarse en la caracterización de tales elementos. Además, los resultados indican que el hecho de que el hablante sepa que no va a ser interrumpido conlleva un incremento de la aparición de hesitaciones en habla espontánea. Los enunciados con pausas y alargamientos se perciben con el mismo grado de naturalidad si se comparan con una versión de los mismos enunciados en la que se han eliminado las hesitaciones, pero no con idéntico nivel de espontaneidad. Finalmente, cabe resaltar que los alargamientos suelen percibirse en el discurso como si el hablante hubiera realizado un silencio.

El objetivo de este artículo es determinar qué puede aportar el conocimiento de los fonetistas al campo de la fonética judicial y cuáles son las diferentes fases que se deben seguir en el análisis de las señales de habla que se emplean en acústica forense. A modo de ejemplo, se aportan algunos datos obtenidos en el marco del proyecto VILE-I (Estudio acústico de la variación inter e intralocutor en español, BFF2001-2551), en el que se han estudiado algunos de los parámetros acústicos que individualizan a un hablante y lo diferencian de los demás.

This paper presents an acoustic analysis of the telephone speaking style in a Brazilian Portuguese corpus. The corpus contains 80 recordings from 20 speakers (10 male and 10 female) in two non-simultaneous conditions: face-to-face (non-mediated speaking style) and via mobile phone (telephone speaking style) and were recorded in silent and noisy environments. The acoustic analysis is divided considering global and local parameters, which are tested with respect to their correlation with condition (face-to-face, mobile phone), environment (silent, noisy) and speakers' gender (male, female). For global parameters, spectral emphasis and F 0-peak production rate in noisy environment tend to be higher in comparison with their values in silent environment. For local parameters, the use of mobile phone in noisy environments causes an increase of the F 0-median. The F 0 standard-deviation is higher in this condition when compared to face-to-face condition. Concerning the differences between genders, F 0-median of male speakers tends to be higher in mobile condition when in noisy environments. Even though females' values for F 0 standarddeviation are higher in each recording condition than males', males' values for this parameter differ more across face-toface vs. mobile phone conditions if compared to women's.

The aim of this research was to investigate the e ciency of a set of acoustic measures to identify one randomly pre-selected individual, named "the criminal", out of a group of ten speakers of Brazilian Portuguese. Among the measures used were the rst two formants of vowels, fundamental frequency, the duration of syllables and vowels, formant movement rate, intensity and the standard deviation of consonantal interval durations (∆C). We analyzed all the Brazilian Portuguese vowels. All the subjects were speakers of Brazilian Portuguese from the states of São Paulo, Rio Grande do Sul, Pará and Bahia. All the samples were extracted from interviews recorded in low quality acoustic environments. The samples were divided into two groups i) interviews and ii) recorded telephone conversations (mobile phone to mobile phone). The parameters that were most robust were rhythm and timing, such as duration, speech rate, ∆C and formant movement rate for the second formant. Because of their interspeaker variability, timing measures proved to be highly discriminative. The statistical tests showed that the speech of three of the subjects contained similarities to that of "the criminal".

У раду су приказани резултати анализе вокалског трајања у изговору 10 мушких говорника косовско-ресавског дијалекта. Фокус истраживања био је на анализи спонтаног говора. Анализиране су једносложне и двосложне речи са наглашеним вокалима који су претходили безвучним плозивима и фрикативима. Изузев броја слогова и фонетског контекста, сагледани су и фактори попут акценатског квантитета, положаја речи у реченици, уједно и темпо говора. Резултати анализе показују да трајање у великој мери зависи од квалитета вокала. Конкретно, најдуже трајање под дугим акцентом имао је ниски вокал средњег реда, док је високи вокал предњег реда у изговору наших испитаника био најкраћи. Међутим, када је реч о вокалима под кратким акцентом, уочена је тенденција испитаника да вокале /е/ и /о/ изговарају са изразито дугим трајањем, што је последица отвореног изговора ових вокала у косовско-ресавском дијалекту. Такође, најмања разлика у трајању вокала под дугим и кратким акцентом забележена је код средњих вокала /е/ и /о/, док je највећa квантитативнa разликa уоченa код високог вокала предњег реда.

У раду су описане спектралне карактеристике самогласника /о/ у спонтаном говору 10 мушких испитаника пореклом са подручја косовско-ресавске дијалекатске области. Дијалектолошки описи говора косовско-ресавског подручја бележе знатно отворенији
изговор вокала /ŏ/ (Окука 2008: 200), те се поменути вокал сматра регионалним маркером за косовско-ресавске говоре (Судимац 2016: 574). Као инструмент за прикупљање корпуса коришћена су питања за дискусију. Анализирани су примери у којима се наглашени вокал /о/ налазио под дугим и кратким акцентом, у звучном и безвучном консонантском окружењу. Приликом обраде снимљеног материјала коришћен је софтверски пакет Praat, верзија 6.2.13 (Бурзма, Вининк 2022). Мерене су вредности прва три форманта (F1, F2, F3), као и трајање вокала. Вредности добијене акустичком анализом су потом
статистички обрађене у програму R (верзија 4.2.1). Резултати спроведене анализе показују да не постоји статистичка значајност у погледу трајања вокала /о/ под дугим акцентом, у поређењу са његовим кратким парњаком, ни у звучном ни у безвучном гласовном окружењу. Што се вокалског квалитета тиче, вокал под кратким акцентом био је отворенији, док у погледу централизације није забележена значајна разлика. У том смислу, резултати тренутне анализе разликују се од резултата претходних студија пажљивог говора, што сугерише да се квалитет кратких и дугих вокала различито манифестује у зависности од говорног стила.

Differentiating subjects through the comparison of their recorded speech is a common endeavor in speaker characterization. When using an acoustic-based approach, this task typically involves scrutinizing specific acoustic parameters and assessing their discriminatory capacity. This experimental study aimed to evaluate the speaker discriminatory power of vowel formants-resonance peaks in the vocal tract-in two different speaking styles: Dialogue and Interview. Different testing procedures were applied, specifically metrics compatible with the likelihood ratio paradigm. Only high-quality recordings were analyzed in this study. The participants were 20 male Brazilian Portuguese (BP) speakers from the same dialectal area. Two speaker-discriminatory power estimates were examined through Multivariate Kernel Density analysis: Log cost-likelihood ratios (C llr) and equal error rates (EER). As expected, the discriminatory performance was stronger for style-matched analyses than for mismatched-style analyses. In order of relevance, F3, F4, and F1 performed the best in style-matched comparisons, as suggested by lower C llr and EER values. F2 performed the worst intra-style in both Dialogue and Interview. The discriminatory power of all individual formants (F1-F4) appeared to be affected in the mismatched condition, demonstrating that discriminatory power is sensitive to style-driven changes in speech production. The combination of higher formants 'F3 + F4' outperformed the combination of lower formants 'F1 + F2'. However, in mismatched-style analyses, the magnitude of improvement in C llr and EER scores increased as more formants were incorporated into the model. The best discriminatory performance was achieved when most formants were combined. Applying multivariate analysis not only reduced average C llr and EER scores but also influenced the overall probability distribution, shifting the probability density distribution towards lower C llr and EER values. In general, front and central vowels were found more speaker discriminatory than back vowels as far as the 'F1 + F2' relation was concerned.

У овом раду испитује се реализација поствокалског апроксиманта /r/ код српских студената енглеског језика у односу на одређене социолингвистичке и лингвистичке факторе. Циљ рада јесте утврдити у којој мери говор српских испитаника одговара ротичком (општеамерички), односно, неротичком (јужни британски стандард) изговорном моделу енглеског језика. Истраживање је спроведено на узорку од 15 испитаника (8 женских и 7 мушких испитаника). Од спољашњих фактора посебно су сагледани: пол, одабрани изговорни модел енглеског језика, као и изложеност испитаника ротичком, тј. неротичком варијетету путем медија. Од унутрашњих фактора испитани су: слоговна структура, квалитет претходног вокала и лексичка фреквентност. Анализа показује да је говор српских  студената знатно ближи (73,99%) ротичком варијетету енглеског језика. Такође, продукција ротичког гласа је у истраживању била чешћа код мушких говорника, посебно оних испитаника који су као свој изговорни модел навели општеамерички варијетет, као и испитаника који су чешће изложени поменутом изговорном моделу. Када су лингвистички фактори у питању, већа учесталост реализације поствокалског /r/ забележена је у медијалној позицији, код нискофреквентних речи и у контекстима где је ротичком гласу претходио вокал задњег реда.

Contemporary technological and educational achievements enable the emergence and development of completely new identification methods based on individual physical characteristics of persons. Today, forensic voice identification is a powerful tool in the fight against crime, in situations where it is necessary to identify a suspect or to acquit an innocent person. Just as there are no two identical fingerprints, two identical retinas, or two identical handwritings, there are no two identical voices. Based on 20-30 seconds of effective speech, forensic experts can perform a voice expertise and identify with high reliability whether a particular voice belongs to a particular person. Comprehensive analysis results in a large amount of data, so by using the statistical analysis method, the necessary forensic markers for forming the final expert opinion are obtained. In this type of expertise, interdisciplinary approach is applied in which several scientific disciplines are included lingu...

This study aimed to analyze the impact of the amount of data on the discriminatory performance of acoustic-phonetic parameters, some of which are frequently assessed in forensic speaker comparisons. Parameters from three distinct phonetic domains were considered, namely, spectral, melodic, and temporal, which were assessed separately within the same phonetic domain and in combination. The speech material consisted of spontaneous telephone conversations between two subjects. During the recording sessions, the participants were placed in different rooms, not directly seeing, hearing, or interacting with each other. The speakers were encouraged to start a conversation using a mobile phone while being simultaneously recorded. All recordings were carried out with a high resolution (44.1 kHz and 16-bit). Data segmentation and transcription were performed in the Praat software [1]. The participants were 20 male subjects, Brazilian Portuguese speakers from the same dialectal area. Their age ranged from 19 to 35 years, with a mean of 26.4 years. Although the subjects (10 identical twin pairs) were recruited from a twin research project, cf. [2, 3, 4], the focus here was comparisons among all speakers (i.e., 190 inter-speaker comparisons) rather than on individual twin pairs. Two metrics of discriminatory performance were examined through the R software [5] as a function of the comparisons among all speakers in the study using the script fvclrr [6]: Log-likelihoodratio-cost (Cllr) and Equal Error Rate (EER) values. The Cllr metric is an empirical estimate of the precision of likelihood ratios. The EER metric captures the point where the false reject rate (type I error) and false accept rate (type II error) are equal and is used to describe the overall accuracy of a system. Lower Cllr and EER values are compatible with better discriminatory performance, whereas higher values suggest the opposite trend. A cross-validation procedure was adopted for the calculation of likelihood ratios, where multiple pairwise comparisons were performed across individuals in which the background sample consisted of data from all speakers, except those being directly compared. To assess the impact of the amount of data on discriminatory performance, a very straightforward approach was adopted. Based on a larger data set extracted from recordings of about 2.50 min per speaker, random data points were additively selected for the analyses. The minimum number of data points, i.e., acoustic measurements, selected per speaker was set at 2 to allow the minimum comparison of intra-speaker variability. Thereafter, more randomly selected data points were progressively added to the tests, two points at a time, and new Cllr and EER values were computed for the new resampled data set. For the present study, the maximum number of data points (acoustic measurements) was set at 30. Given the nature of the speech material assessed, a discrepancy in the number of samples produced per subject was observed. Because of that, a random downsampling procedure was repeated 200 times to minimize the selection bias using the R package Recipes [7]. Cllr and EER median values were reported after performing tests with the randomly selected data while performing a fusion and calibration of different estimates based on a logistic regression technique, cf. [8]. Melodic and temporal parameters were extracted from speech chunks with an average (mean and median) temporal window of 3 s, corresponding to inter-pause intervals in most cases. Spectral parameters were extracted from the midpoints of /a/ vowels in stressed and unstressed positions. The monophthongs displayed a mean and a median duration of 67 ms and 84 ms, respectively. After a manual segmentation, all parameters were extracted automatically using a Praat script [9]. Four models were compared. Model 1 (M1) comprises the combination of melodic parameters (f0 median and f0 base value). Model 2 (M2) comprises the combination of temporal parameters (speech rate and articulation rate). Model 3 (M3) consists of the combination of spectral parameters (F3 and F4). Model 4 (M4) considered the combination of all the acoustic-phonetic parameters.

This pilot study set out to assess the speaker discriminatory power asymmetry regarding parameters from different phonetic dimensions in spontaneous speech, i.e., spectral, melodic, and temporal. The speech material consisted of spontaneous telephone conversations between siblings. The participants were 20 male subjects, Brazilian Portuguese speakers from the same dialectal area. Six acoustic-phonetic parameters were chosen for the comparison: f0 median, f0 baseline, speech rate, articulation rate, F3, and F4. Overall, acoustic parameters pertaining to the speech tempo category depicted the worse performance in terms of speaker discriminatory power when assessed in isolation. Such a trend was indicated by the relatively higher median and mean Cllr and EER values. Moreover, from the set of parameters assessed, high formant frequencies, i.e., F3 and F4, were the bestperforming estimates in terms of discriminability depicting the lowest EER and Cllr values. The results suggested a speaker discriminatory power asymmetry concerning different acoustic-phonetic parameters, in which speech tempo estimates presented a lower discriminatory power when compared to melodic and spectral parameters. The findings also suggest that data sampling is crucial for the reliability of Cllr and EER calculations.

This study aimed to assess what we refer to as the speaker discriminatory power asymmetry and its forensic implications in comparisons performed in different speaking styles: spontaneous dialogues vs. interviews. We also addressed the impact of data sampling on the speaker's discriminatory performance concerning different acoustic-phonetic estimates. The participants were 20 male speakers, Brazilian Portuguese speakers from the same dialectal area. The speech material consisted of spontaneous telephone conversations between familiar individuals, and interviews conducted between each individual participant and the researcher. Nine acoustic-phonetic parameters were chosen for the comparisons, spanning from temporal and melodic to spectral acoustic-phonetic estimates. Ultimately, an analysis based on the combination of different parameters was also conducted. Two speaker discriminatory metrics were examined: Cost Log-likelihood-ratio (Cllr) and Equal Error Rate (EER) values. A general speaker discriminatory trend was suggested when assessing the parameters individually. Parameters pertaining to the temporal acoustic-phonetic class depicted the weakest performance in terms of speaker contrasting power as evidenced by the relatively higher Cllr and EER values. Moreover, from the set of acoustic parameters assessed, spectral parameters, mainly high formant frequencies, i.e., F3 and F4, were the best performing in terms of speaker discrimination, depicting the lowest EER and Cllr scores. The results appear to suggest a speaker discriminatory power asymmetry concerning parameters from different acoustic-phonetic classes, in which temporal parameters tended to present a lower discriminatory power. The speaking style mismatch also seemed to considerably impact the speaker comparison task, by undermining the overall discriminatory performance. A statistical model based on the combination of different acoustic-phonetic estimates was found to perform best in this case. Finally, data sampling has proven to be of crucial relevance for the reliability of discriminatory power assessment.

Identifying speakers by their spoken output is a specialist task for forensic investigators. In the present study we focused on cross-linguistic speaker (Chinese, English, Dutch) identification based on (components of) English stops and fricatives, /p, b, t, d, k, g/ and the fricatives /f, v, θ, ð, s, z, ʃ, ʒ/. The contribution of English noise bursts to native language identification will be presented and the special tokens which contribute the most will be analyzed.

En este trabajo, se presentan los resultados de un estudio en el que se comparan las manifestaciones acústicas del F1 y el F2 de las vocales españolas pronunciadas por seis hablantes de esta lengua como L2 con los formantes vocálicos de estos mismos sonidos articulados por un hablante representativo del español de Concepción (Chile), y también con el F1 y el F2 de las vocales sintetizadas del español descritas por Delattre (1962).

In forensic speaker comparison, it is crucial to decide when completion of the examination may not possible (punt). We explore the factors that make speaker comparison decisions difficult or impossible. These factors may include duration, noise, speaking style, language/dialect, mental state, number of speakers, type and quality of recording, and deception. The analyst needs criteria to decide to reject case work. We present analysis of some of these factors and their impact on automatic speaker recognition systems. We propose a methodology for setting objective thresholds by which comparison examples can be rejected. This methodology could be used by forensic analysts to decide whether or not to proceed with speaker comparisons involving these factors.

Folklorno se stvarala5tvo intenzivno istraZuje u etnologiji, muzikologiji, koreologiji, povijesti umjetnosti i drugim znanostima, a folklorno pjevanje odnedavna i u znanosti o glasu, tj. u uZem podrudju akustici pjevanoga glasa. Heterogenost folklornoga pjevanja koja se

En este trabajo, se presentan los resultados de un estudio en el que se comparan las manifestaciones acústicas del F1 y el F2 de las vocales españolas pronunciadas por seis hablantes de esta lengua como L2 con los formantes vocálicos de estos mismos sonidos articulados por un hablante representativo del español de Concepción (Chile), y también con el F1 y el F2 de las vocales sintetizadas del español descritas por Delattre (1962).

Resumen: VILE (Estudio acústico de la variación inter e intralocutor en español) es un proyecto de investigación básica, financiado por el Ministerio de Ciencia y Tecnología (BFF2001-2551, 2001-2004), cuya finalidad es el estudio acústico de la variación fonética ...

Purpose: to verify contributions of acoustic spectrographic analysis in the forensic identification of speakers with auditorily similar voices, considering the distinctive behavior of acoustic parameters: formants of vowel "é", of connected speech, mean fundamental frequency in Hz, linear prediction curve of vowel "é" and linear prediction curve area; and to propose an objective method to use the analyzed parameters. Methods: a quantitative, qualitative and descriptive study, conducted in Pernambuco on 16 pairs of male siblings, aged 18-60 years. The subjects recorded videos from which the audios were extracted, numbered and sent to three examiners, in two groups: older brothers and younger brothers, for perceptual-auditory pairing. The correct pairings, indicated by at least two examiners, were submitted to acoustic analysis. The statistical tests included Wilcoxon, Kruskal-Wallis and Bonferroni, with p<0.05. Results: the results of analyses of formants and the mean fundamental frequency were not enough to distinguish similar voices. Unprecedentedly, in the measurements of areas generated by the linear prediction curve graphs, a distinctive statistical significance was observed. Conclusion: it was concluded that, among the parameters studied, the measurements of areas of the linear prediction curve objectively indicated effectiveness in distinguishing speakers with auditorily similar voices.

Previous studies have demonstrated the efficacy of speech rhythm measures in speaker identification across various languages with different phonotactic structures. In Persian language, in particular, two categories of speech rhythm metrics were examined: duration-based and intensity-based metrics. Building upon these prior works, the current study delves deeper into the discrimination capabilities of the mentioned measurement types-duration-based versus intensity-based-in the context of Persian speakers. To achieve this, a multinomial logistic regression model was employed on a dataset comprising 20 male Persian speakers, each reciting 100 sentences at a normal speaking pace. Findings revealed that, when distinguishing between Persian speakers, duration-based measures outperform intensity-based ones, however, this excellence is very slight. This observation is significant, as it sheds light on the suitability of specific rhythm metrics for Persian speaker identification. I postulate that this discrepancy in performance may be attributed to the simple syllable structure of Persian and the lesser reliance on intensity as a primary indicator of lexical stress. This research contributes valuable insights into the choice of rhythm metrics for optimal Persian speaker identification and underscores the importance of considering linguistic features when developing speaker recognition systems.

RESUMEN: En este informe científico se exponen completos los resultados finales obtenidos en el proyecto de investigación "IDENTIFICACIÓN PERCEPTIVA DE LOCUTORES PARA LA ACÚSTICA FORENSE" financiado por el Ministerio de Ciencia y Tecnología (DGCYT PB98-0862). Estos resultados fueron publicados parcialmente en Actas del II Congreso de la Sociedad Española de Acústica Forense 1 El objetivo de esta investigación es experimentar y validar un protocolo sistemático para realizar tests de discriminación auditiva en procedimientos de Ruedas de Reconocimiento de Voz (RRV). La investigación se ha realizado siguiendo el método de análisis instrumental 2 y se ha desarrollado en 3 etapas: 1) búsqueda y formalización de las variables fundamentales que actúan en la memoria auditiva para el reconocimiento de una voz; 2) definición de las condiciones óptimas para las variables /duración del discurso/, /dependencia del texto/ y /estado emocional del hablante/ para estimular la memoria auditiva y 3) estudio de la influencia de un periodo de latencia y del estrés emocional en el reconocimiento de voces. Para el desarrollo del estudio se han aplicado diversos métodos de investigación cualitativa y dos procedimientos experimentales.

Kliničko-bolnički centar "Zvezdara", Klinika za otorinolaringologiju, Beograd, Srbija Apstrakt Uvod⁄Cilj. Profesija nastavnika je jedna od vokalno najzahtevnijih profesija. Vokalni zamor odražava se na akustičke karakteristike glasa i dovodi do promene akustičkog kvaliteta tokom vokalne produkcije. Cilj ovog rada bio je da se ispita uticaj vokalnog zamora na akustičke karakteristike glasa kod nastavnika. Metode. Ispitivanje je obuhvatilo 36 nastavnika (16 ispitanika muškog i 20 ispitanika ženskog pola), starosti 27-58 godina. Vokalni zamor pratili smo analizom akustičkih parametara glasa (foniranjem vokala /a/), primenom kompjuterske laboratorije za glas "Kay Elemetrics". Kod ispitanika je sniman glas pre prvog časa, na početku radne nedelje i po završetku poslednjeg časa, na kraju radne nedelje. Obrađeni su parametri signala, šuma i tremora. U obradi podataka primenjene su metode deskriptivne statistike i analitičke statistike (Studentov t-test za zavisne uzorke, χ 2 test, Man Vitnijev U-test i Spirman-ov koeficijent korelacije ranga-ρ). Rezultati. Dobijeni rezultati pokazali su da se statistički značajne razlike u prosečnim vrednostima ispitivanih parametara glasa javljaju kao posledica vokalnog zamora (p < 0,01), dok pol, starost i pušački status nisu imali značajan uticaj na akustičke karakteristike glasa kod naših ispitanika (p > 0,05). Zaključak. Rezultati ovog istraživanja govore u prilog postojanju uzročnoposledične veze između vokalnog zamora i kvaliteta glasa kod vokalnih profesionalaca.

Contemporary technological and educational achievements enable the emergence and development of completely new identification methods based on individual physical characteristics of persons. Today, forensic voice identification is a powerful tool in the fight against crime, in situations where it is necessary to identify a suspect or to acquit an innocent person. Just as there are no two identical fingerprints, two identical retinas, or two identical handwritings, there are no two identical voices. Based on 20-30 seconds of effective speech, forensic experts can perform a voice expertise and identify with high reliability whether a particular voice belongs to a particular person. Comprehensive analysis results in a large amount of data, so by using the statistical analysis method, the necessary forensic markers for forming the final expert opinion are obtained. In this type of expertise, interdisciplinary approach is applied in which several scientific disciplines are included lingu...

Contemporary technological and educational achievements enable the emergence and development of completely new identification methods based on individual physical characteristics of persons. Today, forensic voice identification is a powerful tool in the fight against crime, in situations where it is necessary to identify a suspect or to acquit an innocent person. Just as there are no two identical fingerprints, two identical retinas, or two identical handwritings, there are no two identical voices. Based on 20-30 seconds of effective speech, forensic experts can perform a voice expertise and identify with high reliability whether a particular voice belongs to a particular person. Comprehensive analysis results in a large amount of data, so by using the statistical analysis method, the necessary forensic markers for forming the final expert opinion are obtained. In this type of expertise, interdisciplinary approach is applied in which several scientific disciplines are included lingu...

Different variants of a Forensic Automatic Speaker Recognition (FASR) system based on Emphasized Channel Attention, Propagation and Aggregation in Time Delay Neural Network (ECAPA-TDNN) are tested under conditions reflecting those of a real forensic voice comparison case, according to the forensic_eval_01 evaluation campaign settings. Using this recent neural model as an embedding extraction block, various normalization strategies at the level of embeddings and scores allow us to observe the variations in system performance, in terms of discriminating power, accuracy and precision metrics. From the achieved results it is possible to state that ECAPA-TDNN can be very successfully used as a base component of a FASR system, managing to surpass the previous state of the art, at least in the context of the considered operating conditions.

The identical twins (monozygotic) are siblings (sib) created from the division of one fertilized egg (zygote), so they will be identical in their genetic characteristics and therefore in their phenotypic traits to a very large extent. The voiceprint of the twins is one of these traits. This paper proposed a method to determine the extent of the similarity and dis-similarity between the voiceprints for the brothers of identical twin and thus it is possible to distinguish between their voices. The proposed method adopts the use of the spectrogram as two-dimensional function with neutrosophic transformation. The neutrosophy theory is applied to all spectrogram elements that belong to the first sibling to determine to what extent this element belongs to the spectrogram of the second sibling for the same utterance. This study relied on the use of a number of voice clips collected from 35 identical twins. The neutrosophic transformation is used to convert the voiceprints into the neutrosophic domain represented by three membership functions (true, false, and indeterminate). The results showed that the average extent of the similarity ratio between twins' voices (true membership) is 68.15%, the difference ratio (false membership) is 31.84%, and the indeterminacy membership function ratio is 18.83%.