Speaking Rate Research Articles

Dysarthria associated with traumatic brain injury: speaking rate and emphatic stress

Prosodic abnormality is common in the dysarthria associated with traumatic brain injury (TBI), and adjustments of speaking rate and emphatic stress are often used as steps in treating the speech disorder in patients with TBI-induced dysarthria. However, studies to date do not present a clear and detailed picture of how speaking rate and emphatic stress are affected in this speech disorder. This study, based on the acoustic analyses of syllable repetitions and sentence speech samples, reports on speaking rate and emphatic stress for 12 subjects with TBI and 8 healthy controls. For speaking rate, the subjects with TBI had (1) both slow speaking and articulation rates, (2) smaller phonation proportion and larger pause proportion, and (3) larger percentage change in speaking rate and smaller percentage change in articulation rate. For emphatic stress, the subjects with TBI had (1) significant increases in the difference and percentage change between pre-stressed and pre-unstressed pause durations, (2) significantly smaller difference between stressed and unstressed word durations, but not the percentage change between stressed and unstressed word durations, and (3) significantly reduced differences in f 0 movement and f 0 slope between stressed and unstressed words, but not in RMS range. This study demonstrates the multidimensional nature of prosodic deficits in the dysarthria related to TBI and illustrates the ability of acoustic measures to give a picture of the dysprosody related to TBI-induced dysarthria. Learning outcomes: As a result of this activity, the participant will be able to (1) describe the prosodic disturbances that have been reported in studies of dysarthria associated with TBI; (2) define acoustic measures appropriate to the analysis of changes in speaking rate and emphatic stress; and (3) discuss the importance of prosody to spoken communication.

Intervention Decision-Making for Persons with Amyotrophic Lateral Sclerosis

Intervention Decision-Making for Persons with Amyotrophic Lateral Sclerosis

How to organize a fairly large vowel inventory: the vowels of Fering (North Frisian)

This paper presents parts of the results of the first stage in the phonetic documentation of Fering, a dialect of North Frisian. Descriptions of Fering indicate that it has a large vowel inventory (15 monophthongs, 7 diphthongs, and 3 triphthongs). Ten older male native speakers were recorded, producing the vowel and consonant inventory at different speaking rates, in different phonetic contexts, and in different speaking styles. The results of the acoustic analyses of the vowels of Fering throw light on the questions of a) how the vowel space is exploited in Fering, b) how Fering differentiates its large vowel inventory, c) to what extent consonant–vowel coarticulation affects the acoustic differentiation of vowel categories, and d) to what extent the implementation of the long–short vowel contrasts is affected by speaking rate. The results also provide a more accurate description of the vowel system of an endangered language than do previous auditory accounts.

Intelligibility and preferred rate of Chinese speaking

This research was conducted to investigate the relationship between the rate of natural speech and intelligibility, and to examine the effects that some subject and task factors may have on Chinese speech perception. Other than the objective measure of speech intelligibility, subjective preferences of listeners towards speech rate and presentation mode were also investigated. Sixty subjects performed a set of intelligibility tests and subjective preference evaluations. The results indicated that at the rate of 4.3 characters per second (231 ms per character) previously established for professional newscasters, subjects perceived information with correctness of up to 99% and a maximum preference level was observed. Speech rates and presentation modes that were preferred by the subjects were also more easily intelligible. There was a binaural advantage over the monaural mode of presentation in speech intelligibility, and subjects generally preferred the binaural presentation to the monaural presentation. An accurate prediction of speech intelligibility was possible based on age of listener, presentation mode, and speech rate. The older subjects perceived speech less accurately and preferred slower speech than the younger subjects. The study demonstrated the importance of considering subject and task factors for compilation of standard speaking time for practical applications. Relevance to industry The results of this study are useful in generating a category of basic Chinese speaking times for standardizing speech performance and prediction of speech intelligibility and preference ratings of listeners under different listening conditions. In the reverse, the speaking rate for a certain intelligibility level can be estimated for practical use if detailed listening conditions are known.

Multiregression analysis of autoregressive with exogenous input speech synthesis parameters and voice qualities

This study investigates the relationship between acoustic parameters utilized in the formant-based ARX (autoregressive with exogenous input) speech synthesis model (J. Acoust. Soc. Jpn., 58, 386–397) and perceived voice qualities of synthetic speech. The acoustic parameters manipulated were F0 baseline, F0 range, spectral tilt of glottal flow (TL), formant scaling parameter (FS), and speaking rate (SR). Japanese expressions associated with voice qualities were high-pitched/low-pitched, masculine/feminine, hoarse/clear, calm/excited, powerful/weak, youthful/elderly, thick/thin, and tense/lax (Proc. ICSLP-98, No. 1005). A sentence utterance of an average speaker selected from a database of 109 male speakers was analyzed using the ARX method. Each of the five acoustic parameters of the utterance was manipulated at three levels, producing 243 samples of synthetic speech (3×3×3×3×3). Ten subjects evaluated the voice qualities of each of the 243 synthetic stimuli with regard to the eight Japanese expressions. Multiregression analysis showed that F0 range, F0 baseline, and FS were primary acoustic correlates of high-pitched/low-pitched and masculine/feminine, SR and F0 range for calm/excited, and F0 range, SR and F0 baseline for thick/thin. Significant relations were not found for the remainder of the Japanese expressions, which was thought to be associated in part with irregularities of glottal flow.

The production of the stop voicing distinction in French at normal and rapid speaking rates

This study examined how monolingual Parisian French speakers produced the stop voicing distinction at two different speaking rates (normal, rapid) in two different phonetic contexts (between vowels, between voiceless-fricatives). Results showed that all the acoustic correlates of the voicing distinction under study (syllable-initial: VOT, closure duration and closure voicing; syllable final stops: preceding vowel duration, closure duration, closure voicing and release) were shorter in the rapid speaking rate condition than in the normal speaking rate condition. Furthermore, voicing-related differences were also smaller in the rapid speaking rate condition than in the normal speaking rate condition to a degree that varied with each acoustic correlate and each phonetic context. Overall, similar patterns of voicing-related duration differences and closure voicing differences were found in the normal and rapid speaking rate conditions, but the perceptual salience of small voicing-conditioned duration differences remains to be investigated.

Automatic User-Adaptive Speaking Rate Selection

Today there are many services which provide information over the phone using a prerecorded or synthesized voice. These voices are invariant in speed. Humans giving information over the telephone, however, tend to adapt the speed of their presentation to suit the needs of the listener. This paper presents a preliminary model of this adaptation. In a corpus of simulated directory assistance dialogs the operator's speed in number-giving correlates with the speed of the user's initial response and with the user's speaking rate. Multiple regression gives a formula which predicts appropriate speaking rates, and these predictions correlate (.46) with the speeds observed in good dialogs in the corpus. It is therefore easy, at least in principle, to make systems which adapt their speed to users' needs.

Articulatory characteristics of speakers with apraxia of speech during sentence production at different speaking rates

The present study developed quantitative descriptions of articulatory movements in three speakers with apraxia of speech during sentences produced at different speaking rates. Point-parameterized articulatory data were obtained using the x-ray microbeam (XRMB) technique, and were compared with similar materials from 24 normal young-adult speakers of American English drawn from an existing XRMB database. Speed histories of markers attached to the lips and jaw during production of the test sentence ‘‘The other one is too big.’’ spoken at slow, habitual and fast rates were analyzed to determine mean peak speed, the number of speed peaks, and the timing of particular speed peaks for each sentence replicate by each talker. An attempt was made to derive speed histories for the lip marker expressed relative to concurrent movements of the jaw, to evaluate the relative contribution of jaw speed to lip speed in both talker groups. A description and discussion of results from the analysis will emphasize speaking rate effects on articulator speed and inter-articulator timing.

Effects of speaking rate on the vowel length distinction in Japanese

The issue of acoustic invariance was explored with the question of whether an absolute or relative duration value can be found to reliably classify the Japanese phonemic short and long vowels produced across different speaking rates. The stimuli were Japanese disyllabic non-words (Experiment 1) and real words (Experiment 2). Four native speakers produced these words in a carrier sentence at slow, normal, and fast rates. The duration of accented and unaccented short vowels in CV́CV contexts and contrasting long vowels in CV́VCV and CV́CVV contexts, as well as word duration, was measured. Rate changes affected the duration of long vowels more than short vowels in such a way that the duration difference between short and long vowels was greater for slow than normal, and for normal than fast speech. In contrast with the absolute durations, the ratios of long-to-short vowels and three-mora (CV́VCV or CV́CVV) to two-mora (CV́CV) words were less affected by rate changes. Furthermore, the proportion of the vowel to the total word duration was found to distinguish the two vowel length categories across three rates with high accuracy. The results support the view that “relational” acoustic invariance exists that remains stable across speakers and rates.

Why are non-native (L2) utterances longer than native-produced utterances?

Research has repeatedly shown that utterances produced by non-native speakers are longer than those of native speakers, even when produced as rapidly as possible. The non-native speaking rate effect is known to be greater for late than early L2 learners, but may disappear for late learners when very short-duration L2 utterances are examined. This study examined the duration of English utterances produced by native English (NE) speakers and late native Spanish (NS) learners of English (n=20 each). For normal-rate utterances, durations were significantly longer for the NS than NE group for syntactically complex sentences (e.g., The boy saw that Gary forgot a box of pencils) but not simple sentences (Gary forgot a box of pencils) or noun phrases (a box of pencils). However, the NS-NE difference reached significance for all three utterance types at a fast rate. In Experiment 2, the NS group took longer than the NE group to produce the digit string 1 2 3 4 5 6 7 as rapidly as possible. Taken together, the results suggest that differences in both the articulation of L2 speech sounds and higher-order planning (including the formulation of syntactic representations) contribute to the nonnative speaking rate effect. [Work supported by NIH.]

Compensation for pitch-shifted auditory feedback during the production of Mandarin tone sequences.

Recent research has found that while speaking, subjects react to perturbations in pitch of voice auditory feedback by changing their voice fundamental frequency (F0) to compensate for the perceived pitch-shift. The long response latencies (150-200 ms) suggest they may be too slow to assist in on-line control of the local pitch contour patterns associated with lexical tones on a syllable-to-syllable basis. In the present study, we introduced pitch-shifted auditory feedback to native speakers of Mandarin Chinese while they produced disyllabic sequences /ma ma/ with different tonal combinations at a natural speaking rate. Voice F0 response latencies (100-150 ms) to the pitch perturbations were shorter than syllable durations reported elsewhere. Response magnitudes increased from 50 cents during static tone to 85 cents during dynamic tone productions. Response latencies and peak times decreased in phrases involving a dynamic change in F0. The larger response magnitudes and shorter latency and peak times in tasks requiring accurate, dynamic control of F0, indicate this automatic system for regulation of voice F0 may be task-dependent. These findings suggest that auditory feedback may be used to help regulate voice F0 during production of bi-tonal Mandarin phrases.

Language Model and Speaking Rate Adaptation for Spontaneous Presentation Speech Recognition

The paper addresses adaptation methods to language model and speaking rate (SR) of individual speakers which are two major problems in automatic transcription of spontaneous presentation speech. To cope with a large variation in expression and pronunciation of words depending on the speaker, firstly, we investigate the effect of statistical and context-dependent pronunciation modeling. Secondly, we present unsupervised methods of language model adaptation to a specific speaker and a topic by 1) selecting similar texts based on the word perplexity and TF-IDF measure and 2) making direct use of the initial recognition result for generating an enhanced model. We confirm that all proposed adaptation methods and their combinations reduce the perplexity and word error rate. We also present a decoding strategy adapted to the SR. In spontaneous speech, SR is generally fast and may vary a lot. We also observe different error tendencies for portions of presentations where speech is fast or slow. Therefore, we propose a SR-dependent decoding strategy that applies the most appropriate acoustic analysis, phone models, and decoding parameters according to the SR. Several methods are investigated and their selective application leads to improved accuracy. The combined effect of the two proposed adaptation methods is also confirmed in transcription of real academic presentation.

Toward Spontaneous Speech Synthesis—Utilizing Language Model Information in TTS

State-of-the-art speech synthesis systems achieve a high overall quality. However, synthesized speech still lacks naturalness. To produce more natural and colloquial synthetic speech, our research focuses on integration of effects present in spontaneous speech. Conventional speech synthesis systems do not consider the probability of a word in its context. Recent investigations on corpora of natural speech showed that words that are very likely to occur in a given context are pronounced less accurately and faster than improbable ones. In this paper three approaches are introduced to model this effect found in spontaneous speech. The first algorithm changes the speaking rate directly by shortening or lengthening the syllables of a word depending on the language model probability of that word. Since probable words are not only pronounced faster but also less accurately this approach was extended by selecting appropriate pronunciation variants of a word according to the language model probability. This second algorithm changes the local speaking rate indirectly by controlling the grapheme-phoneme conversion. In a third stage, a pronunciation sequence model was used to select the appropriate variants according to their sequence probability. In listening experiments test participants were asked to rate the synthesized speech in the categories colloquial impression and naturalness. Our approaches achieved a significant improvement in the category colloquial impression. However, no significantly higher naturalness could be observed. The observed effects will be discussed in detail.

Production and perception of Persian geminate stops at three speaking rates

An experiment was designed to determine whether the geminate/singleton category distinction is maintained at fast speaking rates in Persian. Three speakers of Tehrani Persian read test words containing [t,t:,d,d:] in carrier sentences at three speaking rates. The categories do not overlap within a given speaking rate, but the fastest geminates do overlap the normal-rate singletons, implying that the listener must take speaking rate into account in order to perceive the category distinction. The ratio of the consonant closure to the preceding vowel (C/V) is not a useful rate-independent parameter for describing the geminate/singleton boundary in Persian since in Persian the vowel preceding a geminate is slightly longer. However, it was found that the marginal consonant closure (above a minimum closure of about 20 ms) maintains a fixed proportion of the average syllable duration, regardless of rate. This fixed proportion is distinct for geminates and singletons, and so may be used as a single rate-independent parameter for defining the category distinction. Perception tests on natural sentences showed that the distinction is perceptible at each of the three speaking rates. The perceptual response to manipulation of the closure durations indicated that, besides duration, additional cues to the distinction are present.

Talker-specific auditory imagery during reading

The present experiment was designed to determine if auditory imagery during reading includes talker-specific characteristics such as speaking rate. Following Kosslyn and Matt (1977), participants were familiarized with two talkers during a brief prerecorded conversation. One talker spoke at a fast speaking rate and one spoke at a slow speaking rate. During familiarization, participants were taught to identify each talker by name. At test, participants were asked to read two passages and told that either the slow or fast talker wrote each passage. In one condition, participants were asked to read each passage aloud, and in a second condition, they were asked to read each passage silently. Participants pressed a key when they had completed reading the passage, and reading times were collected. Reading times were significantly slower when participants thought they were reading a passage written by the slow talker than when reading a passage written by the fast talker. However, the effects of speaking rate were only present in the reading-aloud condition. Additional experiments were conducted to investigate the role of attention to talker’s voice during familiarization. These results suggest that readers may engage in auditory imagery while reading that preserves perceptual details of an author’s voice.

Effects of Temporal and Spectral Alterations of Speech Feedback on Speech Production by Persons With Hearing Loss

To evaluate the effects of delay and spectral alteration of speech feedback (SF) on the speaking rate and voice pitch in adult users of hearing aids (HAs) and cochlear implants (CIs). Repeated-measure, completely crossed, 2-factor design. Spectral alterations were implemented by replacing SF with noise that was filtered into 1-, 2-, or 4-frequency bands and speech-modulated in real time. Delays varied from 25 to 200 milliseconds. Seven HA users with severe to profound hearing loss and 6 CI users were randomly recruited by public advertising. All were postlingually deafened adults with intelligible speech. The average speaking rate significantly decreased and rate variability significantly increased with increase in SF delay for both groups. Spectral alterations of SF reduced the effect on speaking rate in the HA group but not in the CI group. Spectral alterations did not significantly affect rate variability in the HA group but did so in the CI group. Average voice pitch increased significantly with increase in SF delay and with spectral alterations of SF in the HA group. No significant effects on average pitch of CI users were noted. The 2 groups were affected differently by the delay and spectral alterations of SF. The differences possibly reflect greater spectral resolution ability in the case of CI users and greater audibility of bone-conducted SF (particularly in the low-frequency region) among the HA users.

Word-length and context effects on the acoustics of /ai/

The acoustics of American English /ai/ are investigated. This study investigated whether speakers maintain an invariant slope of the F2 transition across the vowel shortening effects of increased word length and the vowel lengthening effects of following obstruent voicing. Six speakers recorded minimal /ai/ triplets varying in word length and post-vocalic place and voicing (i.e., hide, Heidi, Heidelberg; hype, hyper, hyperness). Analysis of both temporal and spectral characteristics revealed systematic effects of word length and voicing. With increasing word length /ai/ duration significantly decreased across all variables, except in voiceless contexts where there was not a significant decrease from two- to three-syllable words. Before voiced obstruents F2 transition onset increased significantly from one- to two-syllable words with offset frequencies remaining stable. In the three-syllable word condition, the F2 transition offset dropped significantly. Importantly, F2 transition slope remained stable across word lengths, which mirrors earlier findings of constant F2 slope across different speaking rates (Gay, 1968; Miller et al., 1994). However, F2 transition slope was not constant across voicing conditions, being significantly higher before voiceless obstruents than before voiced obstruents. Findings will be compared with earlier work on diphthong acoustics across temporal manipulations.

Feigned depression and feigned sleepiness: A voice acoustical analysis

We sought to profile the voice acoustical correlates of simulated, or feigned depression by neurologically and psychiatrically healthy control subjects. We also sought to identify the voice acoustical correlates of feigned sleepiness for these same subjects. Twenty-two participants were asked to speak freely about a cartoon, to count from 1 to 10, and to sustain an “a” sound for approximately 5 s. These exercises were completed three times (within the same testing session) with three differing sets of instructions to the participants. These three conditions were presented in pseudo-random order to control for any order effects, and all subjects were naı̈ve to the intended purpose of this study. For all three conditions, mean speaking rates and pitch ranges were calculated. A series of paired t tests showed significant differences in the speaking rates (counting numbers and free-speech exercises) between the ‘normal’ and feigned sleepy conditions, and between the normal and feigned depression conditions, but not between the ‘sleepy’ and ‘depressed’ conditions. The results for pitch range, for all speech exercises, were not different between the normal and either the feigned depression or feigned sleepiness conditions. These results indicate that persons feigning depression and sleepiness demonstrate some level of conscious control of their speech rate, but they did not convincingly alter their pitch ranges while feigning depression or sleepiness.

Conversational profiles of children with ADHD, SLI and typical development

Conversational indices of language impairment were used to investigate similarities and differences among children with Attention‐Deficit/Hyperactivity Disorder (ADHD), children with Specific Language Impairment (SLI) and children with typical development (TD). Utterance formulation measures (per cent words mazed and average number of words per maze) differentiated the ADHD group from the SLI and TD groups (ADHD>TD=SLI). In contrast, measures of lexical diversity, average sentence length and morphosyntactic development (number of different words, MLU, and composite tense) differentiated the SLI group from the ADHD and TD groups (SLI<ADHD=TD). High levels of within group variation were observed in children's speaking rate (words per minute). Implications for differential diagnosis and the establishment of phenotypes for developmental language disorders are discussed.

Speaking rates of American and New Zealand varieties of English

Various acoustic measures of speaking rate were calculated for 40 adult speakers of New Zealand English (NZE). These measures were then compared to a group of 40 adult speakers of American English (AE). Results of the analysis identified significantly faster overall speaking rate and articulation rate for the NZE group compared to the AE group. No gender differences in speaking rate or articulation rate were found for either variety of English. The observed differences in rate between AE and NZE are interpreted with regard to diverse vowel production patterns between the varieties, as well as possible differences in rhythmic timing patterns. The impact of differences in speaking rate between varieties of English on the assessment and treatment of speech timing disorders is discussed.