Determining the Likelihood Ratio From Perceptual Attributes of Voice.

PURPOSE: To provide voice experts with a method for determining the likelihood ratio (LR) from the perceptual evaluation of distinctive voice attribute scores. The proposed method aims to obtain the similarity and typicality judgments made by forensic voice experts (FVEs) during the comparison of attributes in voice pairs. METHOD: It is based on the scoring method for LR calculation. In the first stage, 17 perceptual attributes grouped into six vocalic categories are specified. A novel graphical interface is used to obtain discriminative responses both globally and for each attribute from ten pairs of test sentences produced by the same and different speakers. The FVEs should discriminate whether the attributes are similar or different in each pair and should indicate the degree to which the attributes are present. In addition, for six specific attributes, the FVEs must decide whether the attribute is typical or atypical in the reference population. In the second stage, the mean score obtained in the first stage is converted to LR using probability density functions of listeners' responses to 1680 same/different speaker pairs discriminated for female and male speakers. RESULTS: The responses of the FVEs to the test pairs show the discriminatory power of the attributes, the incidence of the typicality factor on the final score and the performance of each FVE. With the application of the probability density functions obtained for the responses to pairs of the same or different origin taken from the reference population, the final scores are converted into LRs that are compared with the true conditions of each pair. CONCLUSIONS AND FUTURE WORK: The application of the developed system allows the global and discriminated evaluation of the perceptual attributes with high agreement in the comparison of pairs of voices. Obtaining the LRs allows associating the perceptual evaluation method with the automatic methods that are used nowadays. The responses of the FVEs taken as a reference, will allow training and evaluating the performance of young FVEs.

A Perceptual Method to Rate Dysphonic Voices.

OBJECTIVE: To present and test a production-matching method with external references, looking at the improvement of inter-rater variability of expert evaluations. METHOD: It consists of adjusting quality attribute levels of a synthetic vowel for a simultaneous matching with the natural patient vowel (NPV) attributes. In an initial experiment, seven speech-language pathology (SLP) experts performed this task with the new method and evaluated the same NPV with the standard method. Targets were twelve NPVs with a variety of quality attribute combinations. In a second experiment, we employed the proposed method to assess the evaluation performance of 65 SLP students. RESULTS: Expert evaluations show less dispersion for the proposed method than those obtained using the standard rating method. Student individual responses were compared with overall responses from their own group and were cross referenced with expert responses. A Kappa index is proposed as a measure of SLP students' performance. CONCLUSIONS: The proposed method was readily accepted by both SLP experts and students. Experts' consensus was improved. SLP students could benefit by quickly learning to discriminate complex attributes, which usually demands years of experience.

Comparison of Two Perceptual Methods for the Evaluation of Vowel Perturbation Produced by Jitter.

OBJECTIVES: To explore perceptual evaluation of jitter produced by fundamental frequency (F0) variation in a sustained vowel /a/, using two different methods. One is based on listener's internal references and the other is based on external references provided by the experimenter. METHODS: We used two methods: one is magnitude estimation-converging limits (ME-CL), which is close to the standard approach used by speech therapists when they use numerical estimations and their own standards, and other is intramodal matching procedure (IMP), where each matched stimulus is to be compared with a fixed-set matching stimuli. Systematic variations were introduced in vowel /a/ by Linear Prediction Coding synthesis using an F0 contour function obtained from a statistical jitter model. Six jitter values were used for each of two reference F0 values. Three groups of listeners were tested: expert speech therapists, speech therapy students, and naïve listeners. RESULTS: Perceptual functions appear to be similar and linear for both methods as the theory predicts. The answers of all groups of listeners tested with ME-CL present higher standard deviations than for IMP. When subjects were tested with IMP, intrareliability and interreliability measurements show a significant improvement for both expert and naïve listeners. CONCLUSIONS: Both intraindividual and interindividual differences for expert speech therapists could be better managed when tested with an IMP than when they use numerical estimations and internal standards to evaluate vowel perturbation produced by jitter. This procedure could be the basis for the development of a clinical evaluation tool.

Índice de perturbación, de precisión vocal y de grado de aprovechamiento de energía para la evaluación del riesgo vocal / Perturbation, precision and extra energy gain indexes for the evaluation of vocal risk

Este trabajo presenta la aplicación de un método adecuado para el análisis del riesgo vocal debido a las alteraciones de voz. Se obtienen 3 índices: a) un índice de perturbación que agrupa 4 parámetros clásicos como el Jitter, el Shimmer, la relación armónico ruido y la amplitud del cepstrum; b) un índice de precisión vocal vinculado con la estabilidad articulatoria y medido como la inversa de la desviación estándar de los primeros 5 formantes, y c) un índice asociado al grado de aprovechamiento de energía que evalúa tanto la coincidencia entre los armónicos con los formantes como las pérdidas de energía que se producen en el tracto vocal, medidas como la inversa de los anchos de banda. Para esta presentación, los índices mencionados se evalúan en 84 voces de docentes con distintos grados de alteración de voz, durante la emisión de la vocal /a/. El índice de perturbación se calcula a partir de las contribuciones parciales sobre una diagonal que va desde valores normales en un extremo hasta valores patológicos en el otro. El índice de precisión vocal se presenta con un gráfico de las áreas de formantes normalizadas respecto de la frecuencia fundamental. El índice de aprovechamiento de energía muestra una gráfica inversa de los anchos de banda a lo largo de un continuo. La agrupación de las voces de docentes en normales, con riesgo vocal y alteradas se presenta en relación a los respectivos diagnósticos laringológicos verificando su utilidad en la evaluación masiva de los profesionales con riesgo vocal (AU)

This study describes the application of an acoustic analysis method that can be used for vocal risk evaluation. Three indexes are calculated: (i) a perturbation index, with classical perturbation parameters, such as jitter, shimmer, harmonic-to-noise ratio, and cepstrum amplitude; (ii) a precision index related to articulatory stability, measured by the inverse of the standard deviation of the first five formants, and (iii) an extra energy gain index due to both optimal harmonic/formant alignment and losses at the vocal tract, measured by the inverse of the first five formant bandwidths. The above indexes were evaluated in 84 school teachers with distinct grades of voice alterations during emission of the vowel /a/. The perturbation index is calculated on the basis of the partial scores along a diagonal line ranging from normal at one end to altered measurements at the other. The precision index is represented as a formant plot - normalized to fundamental frequency - which shows the contributions of both formant one and formant two to the precision index. To verify energy management at the vocal tract, formant bandwidths are measured and their inverse is drawn as points along a non-contribution/contribution line. The teachers’ voices were automatically classified in three groups: normal, at risk and altered, all of which were successfully compared with their laryngologist's diagnosis. The method employed is a promising application for vocal risk evaluation (AU)

Sistemas de análisis acústico y de reconocimiento automático en habla espontánea / Acoustic analysis and automatic recognition systems in spontaneous spech

En este trabajo se presentan dos sistemas de análisis acústico del habla con aplicaciones a la descripción de segmentos de discurso espontáneo y un sistema de reconocimiento automático de habla espontánea orientado a la detección de palabras. El primer sistema de análisis presenta detalladamente todos los rasgos instintivos segmentales y suprasegmentales del habla en forma simultánea asociados a la frecuencia, energía y duración. El segundo presenta automáticamente los parámetros físicos asociados a la entonación en una superficie que cuantifica el campo vocal del hablante y mide el rango vocal y dinámico en el discurso hablado. Se presenta un histograma de la frecuencia fundamental útil para comparar las tendencias entonativas de sesión a sesión. Finalmente se ha desarrollado una herramienta de reconocimiento con modelos acústicos para el español hablado en la Argentina. El mismo transcribe los sonidos grabados a texto y posibilita la aplicación de otras herramientas para el procesamiento de lenguaje natural.

In this paper two acoustic speech analysis systems are presented with applications to the description of spontaneous speech segments and a system of automatic spontaneousspeech recognition oriented to word detection. The first analysis system presents in detail all segment and supra-segment instinct speech features simultaneously and associated frequency, power and duration. The second automatically displays thephysical parameters associated to intonation in a surface that quantifies the vocal field of the speaker and measures the vocal and dynamic range in spoken discourse. A histogram of the fundamental frequency proves useful to compare intonation tendencies from session to session. Finally a recognition tool with acoustic models was developed for Spanish spoken in Argentina. It transcribes the recorded text sounds and enables the application of other tools for natural language processing.

Sistemas de análisis acústico y de reconocimiento automático en habla espontánea / Acoustic analysis and automatic recognition systems in spontaneous spech

En este trabajo se presentan dos sistemas de análisis acústico del habla con aplicaciones a la descripción de segmentos de discurso espontáneo y un sistema de reconocimiento automático de habla espontánea orientado a la detección de palabras. El primer sistema de análisis presenta detalladamente todos los rasgos instintivos segmentales y suprasegmentales del habla en forma simultánea asociados a la frecuencia, energía y duración. El segundo presenta automáticamente los parámetros físicos asociados a la entonación en una superficie que cuantifica el campo vocal del hablante y mide el rango vocal y dinámico en el discurso hablado. Se presenta un histograma de la frecuencia fundamental útil para comparar las tendencias entonativas de sesión a sesión. Finalmente se ha desarrollado una herramienta de reconocimiento con modelos acústicos para el español hablado en la Argentina. El mismo transcribe los sonidos grabados a texto y posibilita la aplicación de otras herramientas para el procesamiento de lenguaje natural.(AU)

In this paper two acoustic speech analysis systems are presented with applications to the description of spontaneous speech segments and a system of automatic spontaneousspeech recognition oriented to word detection. The first analysis system presents in detail all segment and supra-segment instinct speech features simultaneously and associated frequency, power and duration. The second automatically displays thephysical parameters associated to intonation in a surface that quantifies the vocal field of the speaker and measures the vocal and dynamic range in spoken discourse. A histogram of the fundamental frequency proves useful to compare intonation tendencies from session to session. Finally a recognition tool with acoustic models was developed for Spanish spoken in Argentina. It transcribes the recorded text sounds and enables the application of other tools for natural language processing.(AU)

Desarrollo de una prueba de inteligibilidad de habla en ambientes ruidosos para niños en edad escolar / Development of a speech intelligibility test in noisy conditions for children at school age

El objetivo de este trabajo es el diseño y la aplicaciónde una prueba rápida de inteligibilidad del habla enambientes ruidosos. La prueba está diseñada paraniños de 6 a 12 años que asisten a escuelas o instituciones donde no se dispone de equipos de audiometría e instalaciones preparadas para controlar elruido-ambiente. Se evaluará la capacidad de identificarpalabras —mediante auriculares— pronunciadasen el contexto de oraciones en condiciones controladasrespecto del ruido. La prueba se presenta al nivelque requiere cada niño para obtener un reconocimientodel 100 % de las palabras en silencio. Luego,las oraciones y el ruido enmascarador se presentan deforma simultánea en el mismo oído a relaciones señala ruido que van desde 20 a -5 dB. El ruido es el denominado ruido modulado con forma de habla (ModulatedSpeech Shaped Noise, MSSN), que ha demostradoser muy efectivo como enmascarador. Los resultados para niños con audición normal indican 1) el nivel de presión sonora para el reconocimiento del 100 % de las palabras en silencio, medido en el oído artificial es en promedio de 33,9 dB (desviación: 0,9) en el laboratorio y de 36,77 dB (desviación:1,8) en la escuela, y 2) el 50 % de identificación de la palabra se alcanza a la relación señal a ruido de 0 dB tanto en el laboratorio como en la escuela. Se puede concluir que la prueba permitirá evaluaciones rápidas de la inteligibilidad en lugares con diferente nivel de ruido- ambiente y sin disponibilidad de recursos técnicos

A quick intelligibility test for noisy environmentsis presented and evaluated in this article. It isdesigned for children of 6 to 12 years old atschools or institutions where audiometric equipmentsare not easily available. Word identificationcapacity is evaluated in sentences through headphonesunder noise controlled conditions. The testis presented at the level required for each child torecognize 100 % of the words in silence. The sentencesand masking noise are presented simultaneouslyto the same ear at different signal to noiseratios, from 20 to -5 dB. The modulated speechshaped noise was used, which is considered highlyeffective as masker. Results for normal hearingchildren indicate that 1) the sound pressure levelfor 100 % word recognition in isolation, has anaverage of 55 dB and 60 dB in laboratory andschool conditions respectively, and 2) 50 % wordidentification is reached at the 0 db signal to noiseratio, both in the laboratory and the school. It isconcluded that this test will allow quick intelligibilityevaluations in places with different level ofenvironmental noise and absence of technicalresources