Long-Term Average Speech Spectra and Dynamic Ranges of 17 Indian Languages

Purpose: In this work, we have determined the long-term average speech spectra (LTASS) and dynamic ranges (DR) of 17 Indian languages. This work is important because LTASS and DR are language-dependent functions used to fit hearing aids, calculate the Speech Intelligibility Index, and recognize speech automatically. Currently, LTASS and DR functions for English are used to fit hearing aids in India. Our work may help improve the performance of hearing aids in the Indian context. Method: Speech samples from native talkers were used as stimuli in this study. Each speech sample was initially cleaned for extraneous sounds and excessively long pauses. Next, LTASS and DR functions for each language were calculated for different frequency bands. Similar analysis was also performed for English for reference purposes. Two-way analysis of variance was also conducted to understand the effects of important parameters on LTASS and DR. Finally, a one-sample t test was conducted to assess the significance of important statistical attributes of our data. Results: We showed that LTASS and DR for Indian languages are 5–10 dB and 11 dB less than those for English. These differences may be due to lesser use rate of high-frequency dominant phonemes and preponderance of vowel-ending words in Indian languages. We also showed that LTASS and DR do not differ significantly across Indian languages. Hence, we propose a common LTASS and DR for Indian languages. Conclusions: We showed that differences in LTASS and DR for Indian languages vis-à-vis English are large and significant. Such differences may be attributed to phonetic and linguistic characteristics of Indian languages.

Comparing the Informativeness of Single-Word Samples and Connected Speech Samples in Assessing Speech Sound Disorders

Purpose Speech-language pathologists (SLPs) are faced with the challenge of quickly and accurately identifying children who present with speech sound disorders (SSD) compared to typically developing (TD) children. The goal of this study was to compare the clinical relevance of two speech sampling methods (single-word vs. connected speech samples) in how sensitive they are in detecting atypical speech sound development in children, and to know whether the information obtained from single-word samples is representative enough of children's overall speech sound performance. Method We compared the speech sound performance of 37 preschool children with SSD ( M age = 4;11 years) and 37 age-sex-matched typically developing children ( M age = 5;0 years) by eliciting their speech in two ways: (a) a picture-naming task to elicit single words, and (b) a story-retelling task to elicit connected speech. Four speech measures were compared across sample type (single words vs. connected speech) and across groups (SSD vs. TD): intelligibility, speech accuracy, phonemic inventory, and phonological patterns. Results Interaction effects were found between sample type and group on several speech sound performance measures. Single-word speech samples were found to differentiate the SSD group from the TD group, and were more sensitive than connected speech samples across various measures. The effect size of single-word samples was consistently higher than connected speech samples for three measures: intelligibility, speech accuracy, and phonemic inventory. The gap in sample type informativeness may be attributed to salience and avoidance effects, given that children tend to avoid producing unfamiliar phonemes in connected speech. The number of phonological patterns produced was the only measure that revealed no gap between two sampling types for both groups. Conclusions On measures of intelligibility, speech accuracy, and phonemic inventory, obtaining a single-word sample proved to be a more informative method of differentiating children with SSD from TD children than connected speech samples. This finding may guide SLPs in their choice of sampling type when they are under time pressure. We discuss how children's performance on the connected speech sample may be biased by salience and avoidance effects and/or task design, and may, therefore, not necessarily reveal a poorer performance than single-word samples, particularly in intelligibility, speech accuracy, and the number of phonological patterns, if these task limitations are circumvented. Our findings show that the performance gap, typically observed between the two sampling types, largely depends on which performance measures are evaluated with the speech sample. Our study is the first to address sampling type differences in SSD versus TD children and has significant clinical implications for SLPs looking for sampling types and measures that reliably identify SSD in preschool-aged children.

Relative predictive utility of the original and Autism-Specific Five-Minute Speech Samples for child behaviour problems in autistic preschoolers: A preliminary study

Many autistic children have co-occurring behavioural problems influencing core autism symptomology potentially relevant for intervention planning. Parental Expressed Emotion – reflecting critical, hostile and overprotective comments – contributes to understanding and predicting behaviour in autistic school-aged children, adolescents and adults and is typically measured using the Five-Minute Speech Sample. However, limitations exist for its use with parents of younger autistic children and so the Autism-Specific Five-Minute Speech Sample was adapted with the goal of better measuring parent Expressed Emotion in the context of childhood autism. The Autism-Specific Five-Minute Speech Sample has not yet been used to explore Expressed Emotion in parents of autistic preschoolers, nor has the relative predictive utility of the Autism-Specific Five-Minute Speech Sample and Five-Minute Speech Sample been evaluated in the same sample. We compared the two measures from speech samples provided by 51 Australian parents with newly diagnosed autistic preschoolers, including investigating their predictive value for concurrent and subsequent child internalising and externalising behaviour problems. While Autism-Specific Five-Minute Speech Sample Expressed Emotion and Five-Minute Speech Sample Expressed Emotion were associated in this sample, only Autism-Specific Five-Minute Speech Sample codes contributed significant predictive value for concurrent and subsequent child problem behaviour. These preliminary data strengthen the position that the Autism-Specific Five-Minute Speech Sample may better capture Expressed Emotion, than the Five-Minute Speech Sample, among parents of autistic preschool-aged children. Lay abstract Parental Expressed Emotion refers to the intensity and nature of emotion shown when a parent talks about their child, and has been linked to child behaviour outcomes. Parental Expressed Emotion has typically been measured using the Five-Minute Speech Sample; however, the Autism-Specific Five-Minute Speech Sample was developed to better capture Expressed Emotion for parents of children on the autism spectrum. In each case, parents are asked to talk for 5 min about their child and how they get along with their child. Parents’ statements are then coded for features such as number of positive and critical comments, or statements reflecting strong emotional involvement. While both the Five-Minute Speech Sample and Autism-Specific Five-Minute Speech Sample have been used with parents of autistic school-aged children, their relative usefulness for measuring Expressed Emotion in parents of preschool-aged children – including their links to child behaviour problems in this group – is unclear. We collected speech samples from 51 parents of newly diagnosed autistic preschoolers to investigate similarities and differences in results from the Five-Minute Speech Sample and Autism-Specific Five-Minute Speech Sample coding schemes. This included exploring the extent to which the Five-Minute Speech Sample and Autism-Specific Five-Minute Speech Sample, separately, or together, predicted current and future child behaviour problems. While the two measures were related, we found only the Autism-Specific Five-Minute Speech Sample – but not the Five-Minute Speech Sample – was related to child behavioural challenges. This adds support to the suggestion that the Autism-Specific Five-Minute Speech Sample may be a more useful measure of parental Expressed Emotion in this group, and provides a first step towards understanding how autistic children might be better supported by targeting parental Expressed Emotion.

Speech Emotion Recognition Using MLP Classifier

Language is human's most important communication and speech is basic medium of communication. Emotion plays a crucial role in social interaction. Recognizing the emotion in a speech is important as well as challenging because here we are dealing with human machine interaction. Emotion varies from person to person were same person have different emotions all together has different way express it. When a person express his emotion each will be having different energy, pitch and tone variation are grouped together considering upon different subject. Therefore the speech emotion recognition is a future goal of computer vision. The aim of our project is to develop the smart emotion recognition speech based on the convolutional neural network. Which uses different modules for emotion recognition and the classifier are used to differentiate emotion such as happy sad angry surprise. The machine will convert the human speech signals into waveform and process its routine at last it will display the emotion. The data is speech sample and the characteristics are extracted from the speech sample using librosa package. We are using RAVDESS dataset which are used as an experimental dataset. This study shows that for our dataset all classifiers achieve an accuracy of 68%.

Rodgers, Lau, & Zebrowski (preprint) - "Examining the effects of stuttering and social anxiety on interpretations of ambiguous social scenarios among adolescents"

PURPOSE: The proclivity to construe ambiguous information in a negative way is known as interpretation bias, which has been implicated in the onset and/or maintenance of social anxiety. The purpose of this study was to examine group and individual differences in interpretation bias among young people who stutter and their typically fluent peers during the adolescent years when social fears and worries tend to escalate. METHODS: A total of 99 adolescents (13 to 19 years old) participated, including 48 adolescents who stutter (67% male) and 51 typically fluent controls (68% male). They completed a computerized vignette-based interpretation bias task in which they first read 14 short ambiguous social scenarios (half including a verbal interaction, half including a non-verbal interaction). They were then presented with four possible interpretations of each scenario including two negative interpretations (one target, one foil) and two positive interpretations (one target, one foil). Participants used a 4-point Likert scale to rate how similar in meaning each interpretation was to the original scenario. Participants also completed self-report measures of social and general anxiety, and provided a speech sample for stuttering analysis. RESULTS: There was no effect of stuttering on interpretations; the adolescents who stutter rated interpretations across both verbal and non-verbal scenarios comparably to the controls, and stuttering severity did not affect interpretation ratings. However, across groups, there was a significant effect of social anxiety such that higher social anxiety was associated with more negative interpretations, and lower social anxiety was associated with more positive interpretations.DISCUSSION: This study provides preliminary evidence that social anxiety may affect how adolescents interpret ambiguous social cues in verbal and non-verbal scenarios more than stuttering, although more research into how people who stutter process social information is warranted.

Performance of Forced-Alignment Algorithms on Children's Speech

Purpose Acoustic measurement of speech sounds requires first segmenting the speech signal into relevant units (words, phones, etc.). Manual segmentation is cumbersome and time consuming. Forced-alignment algorithms automate this process by aligning a transcript and a speech sample. We compared the phoneme-level alignment performance of five available forced-alignment algorithms on a corpus of child speech. Our goal was to document aligner performance for child speech researchers. Method The child speech sample included 42 children between 3 and 6 years of age. The corpus was force-aligned using the Montreal Forced Aligner with and without speaker adaptive training, triphone alignment from the Kaldi speech recognition engine, the Prosodylab-Aligner, and the Penn Phonetics Lab Forced Aligner. The sample was also manually aligned to create gold-standard alignments. We evaluated alignment algorithms in terms of accuracy (whether the interval covers the midpoint of the manual alignment) and difference in phone-onset times between the automatic and manual intervals. Results The Montreal Forced Aligner with speaker adaptive training showed the highest accuracy and smallest timing differences. Vowels were consistently the most accurately aligned class of sounds across all the aligners, and alignment accuracy increased with age for fricative sounds across the aligners too. Conclusion The best-performing aligner fell just short of human-level reliability for forced alignment. Researchers can use forced alignment with child speech for certain classes of sounds (vowels, fricatives for older children), especially as part of a semi-automated workflow where alignments are later inspected for gross errors. Supplemental Material https://doi.org/10.23641/asha.14167058

IDENTIFICATION OF FAMILIAR WORDS FOR HINDI SOUNDS

The Speech Language Pathologists and language experts need material to collect the speech sample which they can evaluate and analyze for normalcy. For older children the speech sample can be collected even in spontaneous speech or by reading of standardized text, but this cannot be done for younger children who cannot read sentences and words. For these children standardized set of word list is required so that their phonology can be checked for normalcy and intelligibility. This word list must not only be structured for presence of each sound at all positions but also these words should be familiar to the younger age group (present in their vocabulary) as the need to identify a picture for it and name it. Such structured material is still limited in Hindi Language. The present study aims the development of word list in Hindi Language and check the familiarity of the word list. The word list prepared was shown to 10 teachers of preschool (Nursery to Upper Kindergarten). The words were rated on a three point rating scale and the results were analyzed using descriptive statistics. Those words found more than 75% familiarity may be used with younger children for speech sample collection. The words with familiarity between 50 to 75% can be used with younger children along with few semantic and phonetic cues.

Teaching English pronunciation to a Russian speaker (a single-case analysis)

This article provides an analysis of a Russian-speaking student's speech sample compared to British normative pronunciation. A number of inconsistencies with the "generally accepted pronunciation" of vowels and consonants,as well as the rhythm of speech, were identified in the speaker's pronunciation. The article discusses the most common mistakes in the student's speech and offers phonetic exercises to improve articulation.

VOTIC-INGRIAN CONVERGENCE AND INTRA-IDIOLECTAL CONTINUUM (A CASE STUDY OF A CHAIN RUNE)

This paper analyses different variants of the Votic chain rune Kuza piippu? “Where is the pipe?” in the context of Votic-Ingrian convergent processes. The main focus is made on the alternation between the lexemes “granary” and “fence”, and the structure of postpositional phrases containing these lexemes. The analysis is based on 13 variants of the rune published by several researchers, and three variants of the same rune recorded by the author in the village of Luuditsa of the Kingisepp region. In different variants of Kuza piippu?, three lexemes alternate within the same line: ratiz ‘granary’, aitta ‘granary’, and aita ‘fence’. The paper concludes that the first variant is the original Votic lexeme meaning ‘granary’, the second one is an Ingrian word that was not fully adopted by Votic, and the third variant emerged as a substitution of the unfamiliar Ingrian word with the phonetically closest Votic word. The Ingrian influence is observed also in the postpositional phrase with the discussed lexemes (‘under the granary ~ fence’). In the earlier versions of the rune, one finds the postposition alla ‘under’ as a separate word. In more recent variants, the head noun and postposition are usually written together as one word, with a formative n between them. This n is the Ingrian marker of the genitive case that was later re-analyzed as the initial consonant of the postposition (alla > nalla). The research has revealed that even in the variants recorded from the same speaker, the combination of Votic and Ingrian elements is almost arbitrary. The Votic-Ingrian ratio is not as much a characteristic of the idiolect, but rather a characteristic of a particular text. Therefore, the idiolect cannot be considered as a minimal sociolinguistic object. The author introduces the notion of “variolect” as a language variant with a particular ratio of languages in contact that characterizes a given speech sample. The mixing of Votic and Ingrian in the western Votic villages is a vivid example of iterative convergence. The Lower Luga Ingrian that emerged as a convergent variety on the basis of several Finnic languages (Ingrian and Votic, most of all), gives birth to new contact varieties when acquired by Votic speakers.

The Relationship Between Speech Characteristics and Motor Subtypes of Parkinson's Disease

Purpose The aim of this study was to examine whether acoustic dysarthria characteristics align with overall motor profile in individuals with Parkinson's disease (PD). Potential speech differences between tremor-dominant and non–tremor-dominant subtypes are theoretically motivated but empirically inconclusive. Method Twenty-seven individuals with dysarthria from PD provided a contextual speech sample. Participants were grouped into non–tremor-dominant ( n = 12) and tremor-dominant ( n = 15) motor subtypes according to the Unified Parkinson Disease Rating Scale. Dependent speech variables included fundamental frequency range, average pause duration, cepstral peak prominence, stuttering dysfluencies, and maze dysfluencies. Results There were no significant differences between the speech of the tremor-dominant and non–tremor-dominant groups. High within-group variability existed across parameters and motor subtypes. Conclusion Speech characteristics across the areas of phonation, prosody, and fluency did not differ appreciably between PD motor subtypes.