Word Intelligibility Research Articles

Reexamining Synthetic Speech: Intelligibility and the Effects of Age, Task, and Speech Type on Recall

Synthetic speech is a technology that can be utilized to convey information and aid people in their tasks. Older adults in particular are a population that may be able to benefit from synthetic speech, and they are a population that has been investigated in a limited capacity. The current researchers intended to elucidate lingering conflicts in previous research on the intelligibility and recall of word and stories in synthetic speech for older and younger adults and how that compared to similar conditions in natural speech. Twenty-four older and 24 younger adults completed intelligibility and recall tasks with word lists and stories. Results indicated that older adults had a more difficult time with all speech, natural speech was easier to understand and remember than synthetic speech, and stories were easier to recall than words. Results also indicated that older adults had a more difficult time understanding synthetic words as compared to natural words than younger adults. In addition, older adults improved differentially with the recall of stories as opposed to words when compared to the younger adult group. Potential directions for synthetic speech software design and future research are discussed.

Acceptable range of speech level for both young and aged listeners in reverberant and quiet sound fields

The speech level of verbal information in public spaces should be determined to make it acceptable to as many listeners as possible, while simultaneously maintaining maximum intelligibility and considering the variation in the hearing levels of listeners. In the present study, the universally acceptable range of speech level in reverberant and quiet sound fields for both young listeners with normal hearing and aged listeners with hearing loss due to aging was investigated. Word intelligibility scores and listening difficulty ratings as a function of speech level were obtained by listening tests. The results of the listening tests clarified that (1) the universally acceptable ranges of speech level are from 60 to 70 dBA, from 56 to 61 dBA, from 52 to 67 dBA and from 58 to 63 dBA for the test sound fields with the reverberation times of 0.0, 0.5, 1.0 and 2.0 s, respectively, and (2) there is a speech level that falls within all of the universally acceptable ranges of speech level obtained in the present study; that speech level is around 60 dBA.

The effect of levodopa on respiration and word intelligibility in people with advanced Parkinson's disease

Objectives Hypokinetic dysarthria is commonly encountered in Parkinson's disease (PD). Although the etiology of this dysarthria is multifactorial, disorders of respiration may strongly contribute to it. The aim of this study was to evaluate the effects of levodopa on measures of respiration and intelligibility. Patients and methods Vital capacity (VC), sustained vowel phonation (SVP) and phonation quotient (PQ) were determined with and without medication in 25 patients suffering from idiopathic PD. Intelligibility was evaluated by three independent speech pathologists using the word subtest protocol of the Yorkston and Beukelman “Assessment of intelligibility of dysarthric speech” (AIDS). Results VC was abnormal in 18/25 patients in the off-condition and 15/25 in the on-condition. SVP was normal in 22/25 patients in the off-state and in all patients in the on-state. PQ was normal in all patients in the off- and on-condition. All three respiratory parameters, as well as intelligibility, improved significantly following administration of levodopa. There were significant differences between men and women for VC and PQ, with men having the highest values for both parameters in both conditions. In none of the two conditions a correlation between respiratory parameters and intelligibility could be demonstrated. Conclusion This study suggests that thoracic mobility is decreased in PD, and that pharmacological treatment results in improvement, but not in normalization. Moreover, it remains unclear to what extent dyskinesias negatively influence respiratory control. Our results also demonstrate improvement of intelligibility after pharmacological treatment of PD. However, this improvement is not solely the consequence of respiratory changes. Further research on the effects of different characteristics on intelligibility is necessary.

Effects of aging on word intelligibility and listening difficulty in various reverberant fields

Listening difficulty ratings is a new measure to evaluate speech transmission performance in public spaces. The advantage is that it can evaluate sound fields with higher speech transmission performance more accurately and sensitively than intelligibility scores. The effect of aging, or hearing level on listening difficulty ratings has not been clarified yet. In this paper, the results of the listening tests demonstrated that hearing levels above 2 kHz significantly affect both word intelligibility scores and listening difficulty ratings for the aged. For the aged, both of them can be estimated from the results for the young plus the average of hearing levels at 2 and 4 kHz. The equations for estimating the word intelligibility scores and the listening difficulty ratings for the aged are obtained from the multiple linear regression analyses. The estimated scores and ratings generally coincide with the measured scores and ratings, and the averaged difference between them is 5.2% for the word intelligibility scores and 3.8% for the listening difficulty ratings, in the range of listening tests performed in this paper.

Behavioural intervention effects in dysarthria following stroke: communication effectiveness, intelligibility and dysarthria impact

Dysarthria is a common post-stroke presentation. Its management falls within the remit of the speech and language therapy profession. Little controlled evaluation of the effects of intervention for dysarthria in stroke has been reported. The study aimed to determine the effects of a period of behavioural communication intervention on communication effectiveness and intelligibility, and of the speech disorder's impact on the person with dysarthria. Eight people with dysarthria following stroke received individually tailored intervention programmes of 16 sessions during an 8-week period. Conversation, and reading aloud of connected speech and of single words were sampled at four points, at intervals of around 2 months: two before intervention and two after intervention. The data were perceptually evaluated by ten listeners, blind to assessment point, for the overall effectiveness of communication (conversation) and for single-word and reading intelligibility. The impact of dysarthria was assessed at the beginning and end of the intervention period. The level of reliability of listener ratings was high. There was no evidence of group change between assessment points for measures of communication effectiveness during conversation, or for reading and word intelligibility. Individual participant analysis indicated that five speakers improved in at least one of these three measures. In all cases this followed either stability or deterioration during the 2-month pre-intervention period. There was very little evidence of deterioration during the 2 months after withdrawal of treatment. For three participants no intervention-related speech benefit was demonstrated. Age, dysarthria severity or lesion information did not appear to differentiate those who improved from those who did not. Improvement occurred in all who began the intervention phase between 5 and 8 months following stroke onset. Group data indicated no change in the Dysarthria Impact Profile in respect of three sections (The effect of dysarthria on me as a person; How I feel others react to my speech; and How dysarthria affects my communication with others). There was a significant difference between the start and end of intervention for Accepting my dysarthria, suggesting a reduction in the negative impact of dysarthria. The results demonstrate that some individuals with dysarthria have a capacity to respond positively to intervention, some months after stroke, and to maintain this improvement following 2 months of no intervention. Consideration is given to how the results of the present study may inform subsequent phases of dysarthria stroke research.

Technocultural education

In this essay I try to describe the development of a traditional liberal education into a technological liberal one. I propose that we start by dropping the classical oppositions between man and animal, and man and machine; that we stop pitting morality against technology and rhetoric; and that we do away with the idea that ICT in our schools will necessarily tear the fabric of education apart. We should rather try and re-describe the idea of an unencumbered and independent self in terms of relational concepts, like the cyborg or more radically: like the self as interface. John Dewey led the way to this view a century ago, by coining the word intelligence as the name of educative interactions between man, animal and machine. The self as interface is a self of differences rather than identities. But that idea does not do away with our emplaced body, or our personal sense of self and identity. In the postmodern world, the cyborg is a migrant with the ability to interpret signs, understand symbols of power, see through rhetorical games, engage in argumentation, and in these activities partake in his or her own political education. The Internet nomad does not bode anarchy. He or she is the radically decentred subject that may well participate in Kant’s cosmopolitanism, Jürgen Habermas’ discourse ethics and Jacques Derrida’s deconstruction. – But this is to go slightly beyond the text submitted here …

Effects of reproduction methods and audio coding on word intelligibility with competing talkers

Effects of reproduction methods and high-efficiency audio coding on word intelligibility were examined in a tele-collaboration system. Monaural, stereophonic, and binaural methods were studied. Transform domain weighted interleave vector quantization (TwinVQ) [Iwakami et al., Proc. ICASSP’95, pp. 937–940 (1995)] was used as the audio coding method. Intelligibility tests using word lists, in which word familiarity was controlled, were conducted under six conditions combining three reproduction methods with two conditions, i.e., with or without audio coding. In the experiments, three talkers pronounced different words simultaneously; a listener was then asked to selectively hear the word of a designated talker. Analyses of variance of the experimental results showed that intelligibility for the binaural method was higher than for the other two reproduction methods in both conditions with or without audio coding. Moreover, the benefit of the binaural method was demonstrated because intelligibility for the method with audio coding was higher than that for the stereophonic method without coding, whereas the amount of audio data in the former method was much smaller than in the latter, indicating that the binaural method with high-efficiency audio coding is useful for a tele-collaboration system. [Work supported by JSPS: KAKENHI (17560336) and RIEC, Tohoku Univ. (H18/A10).]

Development of word lists for word intelligibility tests in Japanese

Speech intelligibility tests are widely used to evaluate personal speech hearing ability accurately. Although various lists have been proposed for the tests, almost all lists in Japan do not properly consider word difficulty as an aspect of speech recognition. Consequently, even if these tests are performed under identical conditions, correct results cannot be obtained. We have been developing word lists for word intelligibility tests to cope with these problems. We control word familiarity, which strongly affects word recognition scores, to equalize difficulty in recognition between word lists. Words are divided into four word-familiarity ranks. Twenty lists of 50 words are constructed in each rank by considering phonetic balance. Moreover, to compensate slight differences in intelligibility scores among word lists within the same word-familiarity rank, the relationship between the difference of word familiarity and signal-to-noise ratio (SNR) is estimated. Then this relationship is used to compensate the difference of the intelligibility score. The other compensation method is based on speech recognition threshold (SRT). The intelligibility score of each word can be equalized by changing the SNR according to the difference of the SRT of the word. The effectiveness of these compensation methods is discussed.

Relationships between the intelligibility and acoustic properties of native English and Mandarin-accented English

Studies examining effects of talkers’ and listeners’ linguistic backgrounds on speech intelligibility have produced mixed results, and relatively little is known about which acoustic-phonetic features of native and non-native speech are responsible for many findings. These issues are addressed here via investigation of how acoustic properties of final consonant voicing contrasts influence word intelligibility. Native English and native Mandarin talkers’ productions of English words were presented in isolation to native English and native Mandarin listeners in an identification task. Intelligibility findings were related to preceding vowel duration, stop consonant duration, burst duration, and proportion of voicing during stop closure. As expected, native English listeners were more accurate identifying words spoken by native than non-native talkers. In contrast to previous findings of an interlanguage speech intelligibility benefit (ISIB) for sentences, non-native listeners found native speech more intelligible than non-native speech. However, non-native listeners found non-native speech more intelligible than did native listeners. For native talkers, all four acoustic measures contributed significantly to correct identification of voicing contrasts by both native and non-native listeners (with closure voicing most salient). In contrast, for non-native talkers, listeners tended to rely primarily on closure voicing and burst duration. Ramifications for ISIB will be discussed.

Word and mora intelligibility in ‘‘Familiarity-controlled word-lists 2003 (FW03).’’

‘‘Familiarity-controlled Word-lists (FW03)’’ was developed to make it possible to perform intelligibility tests on the same person repeatedly and/or under several different conditions. FW03 consists of 20 lists of 50 words in four word familiarity ranks. These words were selected taking phonetic balance into consideration to maximize their variety of initial moras and vowel-consonant sequences. To confirm that all FW03 word lists from a particular familiarity rank present the same hearing difficulty, we measured the word intelligibility of all the words in FW03. Sixteen subjects listened to 4000 words spoken by four speakers with seven signal-to-noise ratios. The mean intelligibility scores for the lists were significantly different even for the same familiarity rank. The intelligibility was apparently influenced by some specific moras and their position in words. These results suggest that it is difficult to equalize the word intelligibility of the lists even when the word familiarity is controlled more precisely. We therefore examined several methods for equalizing the intelligibility among the lists such as controlling the sound pressure level for each word when operating an actual intelligibility test. The effectiveness of these methods is discussed with detailed analyses of word and mora intelligibility in FW03.

Effects of Background Noise on Communication in a Collaborative Team Environment

Many command and control operations involve complex collaborative tasks that cannot be successfully completed without effective team communication. However, the ability to communicate efficiently in these tasks is often compromised by environmental factors, such as the presence of ambient noise. In this study, we attempted to quantify the effect that ambient noise might have on the completion of team tasks in an airborne conference room, where there is a severe cost and weight penalty for the noise countermeasures used to achieve each additional decibel of attenuation. Rather than relying on current ambient noise standards, which are based solely on point-to-point word intelligibility, we chose to base our measurement on a collaborative team “communicability” task that required subjects to achieve a consensus on the ordering of four random faces in each of eight labeled columns. Completion time in the task was measured with and without visual line-of-sight in three levels of ambient noise (55dBA, 70dBA & 80dBA). The results show that completion time in the task increased systematically with the level of the noise, but that visual cues had no significant effect on performance. The results are discussed in terms of their impact on the interpretation of the current standard for ambient noise in communication environments, and in terms of their more general implications for the study of team communication.

Intensive dysarthria therapy for older children with cerebral palsy: findings from six cases

Children with cerebral palsy often have speech, language and communication difficulties that affect their access to social and educational activities. Speech and language therapy to improve the intelligibility of the speech of children with cerebral palsy has long been advocated, but there is a dearth of research investigating therapy effectiveness. To begin to explore the possible effectiveness of therapy focussing on speech production, six students aged 10-18 years, were given intensive individual therapy to increase breath support for speech, maintain breath support and volume across utterances and hence maximise their background effort for speech. Intelligibility of single words and connected speech was measured immediately before and after the block of therapy and again seven weeks later. Gains in intelligibility were observed at the single word level, but not for continuous speech. Implications for therapy and future research are discussed.

Making non-fluent aphasics speak: sing along!

A classic observation in neurology is that aphasics can sing words they cannot pronounce otherwise. To further assess this claim, we investigated the production of sung and spoken utterances in eight brain-damaged patients suffering from a variety of speech disorders as a consequence of a left-hemisphere lesion. In Experiment 1, the patients were tested in the repetition and recall of words and notes of familiar material. Lyrics of familiar songs, as well as words of proverbs and prayers, were not better pronounced in singing than in speaking. Notes were better produced than words. In Experiment 2, the aphasic patients repeated and recalled lyrics from novel songs. Again, they did not produce more words in singing than in speaking. In Experiment 3, when allowed to sing or speak along with an auditory model while learning novel songs, aphasics repeated and recalled more words when singing than when speaking. Reduced speed or shadowing cannot account for this advantage of singing along over speaking in unison. The results suggest that singing in synchrony with an auditory model--choral singing--is more effective than choral speech, at least in French, in improving word intelligibility because choral singing may entrain more than one auditory-vocal interface. Thus, choral singing appears to be an effective means of speech therapy.

Accuracy of Repetition of Digitized and Synthesized Speech for Young Children in Background Noise

The present study investigated the intelligibility of digitized and synthesized speech output in background noise for children 3-5 years old. The purpose of the study was to determine whether there was a difference in the intelligibility (ability to repeat) of 3 types of speech output (digitized, DECTalk synthesized, and MacinTalk synthesized) in single words and sentences, presented within and out of context. The dependent variable was speech intelligibility (number of individual words repeated correctly). The study used a mixed-model design. Ninety typically developing children (3-5 years old) were assigned to each of 3 speech type conditions. Participants were asked to repeat 20 words and 10 short sentences. Half of the stimuli were preceded by contextual information (topic cue), and half were presented without any context. Young children have difficulty accurately repeating some digitized and synthesized messages in background noise. Overall, the older children (4- and 5-year-olds) performed better than the 3-year-old children. Increasing information through context or longer messages (i.e., sentences) did facilitate intelligibility overall, although there was a statistically significant Message Length x Context x Speech Type interaction. For 3-5-year-olds, the intelligibility of single words is very low (55%-77%). The intelligibility of sentences is higher, but the sole use of sentences for communication is problematic. Contextual information facilitates intelligibility and is a promising approach for ensuring effective communication. Future research is needed to improve the intelligibility of speech output at the single word level in order to maximize the benefits of speech output.

Vowel errors and sentence intelligibility in Spanish-accented English speech

It has been shown that foreign-accented speech is more difficult for listeners to process than native speech when presented under noisy listening conditions. It has also been shown that errors in vowel productions are more damaging to connected speech intelligibility than are errors or distortions in consonant production. This study compared the intelligibility of Spanish-accented English vowels in isolated words with that of short samples of connected speech. Five native Spanish speakers who had learned English as adults read a list of 42 monosyllabic English words selected so that 14 Spanish-accented English monophthongs could be presented to native listeners in a three-alternative forced-choice identification task. In each trial the target and two nearby (in the vowel space) distracters were presented. Talkers also read a set of 20 phonetically balanced sentences. Ten native English listeners identified the vowels in the isolated word task and judged the intelligibility of the sentences in a simple write-down procedure. While overall sentence intelligibility scores are not very well predicted by word intelligibility scores when the full set of vowel contrasts is included, predictability improves when a subset of the vowels is chosen. [Work supported by Purdue Research Foundation.]

Evaluation des Oldenburger Kinder-Reimtests in Ruhe und im Störgeräusch

The Oldenburg children's rhyme test (OlKi) was designed and optimized for speech intelligibility measurements for primary school pupils in silence [8, 3]. In the optimization, the intelligibility of the particular test words was equalized. The evaluation of the test with 147 primary school pupils with normal hearing in silence and 107 pupils in noise is presented in this article. The comparability between test lists for speech intelligibility was investigated and age dependent reference functions were determined. The evaluation showed that intelligibility differences are larger across children within one grade than across different test lists. The reference functions of first grade pupils are shifted to slightly higher presentation levels and signal-to-noise ratios both in silence and noise. The 12 optimized test lists of the OlKi test are equally intelligible both in silence and noise. No list effects are expected. The degree of difficulty of the Oldenburg children's rhyme test can be compared with the Göttingen children's test II and the Mainz children's test III.

Contribution of pitch-accent information to Japanese spoken-word recognition

This paper investigates the contribution of pitch-accent information to Japanese spoken-word recognition. Pitch accent of spoken words was manipulated by controlling F0. First, the present author investigated the relation between word intelligibility in the presence of noise and the adequacy of accent type in those words. In the intelligibility test, participants were presented with speech stimuli and a pink noise together, and were required to identify the word. In the rating test, the same participants were presented with the same speech stimuli and were required to rate the adequacy of the words’ accent types. Results indicated that the spoken words with more adequate accent type were more intelligible in the presence of noise. Next, the present author investigated the relation between reaction time in shadowing words and the adequacy of accent type in those words. In the shadowing task, the participants were required to shadow a word whose accent type was manipulated as soon as they identified it. The same participants participated in the rating test. The reaction time in the case of the words with an adequate accent was shorter than in the case of an inadequate one. These results support the hypothesis that pitch-accent information in Japanese spoken words might facilitate word recognition.

Effects of Instantaneous Multiband Dynamic Compression on Speech Intelligibility

The recruitment phenomenon, that is, the reduced dynamic range between threshold and uncomfortable level, is attributed to the loss of instantaneous dynamic compression on the basilar membrane. Despite this, hearing aids commonly use slow-acting dynamic compression for its compensation, because this was found to be the most successful strategy in terms of speech quality and intelligibility rehabilitation. Former attempts to use fast-acting compression gave ambiguous results, raising the question as to whether auditory-based recruitment compensation by instantaneous compression is in principle applicable in hearing aids. This study thus investigates instantaneous multiband dynamic compression based on an auditory filterbank. Instantaneous envelope compression is performed in each frequency band of a gammatone filterbank, which provides a combination of time and frequency resolution comparable to the normal healthy cochlea. The gain characteristics used for dynamic compression are deduced from categorical loudness scaling. In speech intelligibility tests, the instantaneous dynamic compression scheme was compared against a linear amplification scheme, which used the same filterbank for frequency analysis, but employed constant gain factors that restored the sound level for medium perceived loudness in each frequency band. In subjective comparisons, five of nine subjects preferred the linear amplification scheme and would not accept the instantaneous dynamic compression in hearing aids. Four of nine subjects did not perceive any quality differences. A sentence intelligibility test in noise (Oldenburg sentence test) showed little to no negative effects of the instantaneous dynamic compression, compared to linear amplification. A word intelligibility test in quiet (one-syllable rhyme test) showed that the subjects benefit from the larger amplification at low levels provided by instantaneous dynamic compression. Further analysis showed that the increase in intelligibility resulting from a gain provided by instantaneous compression is as high as from a gain provided by linear amplification. No negative effects of the distortions introduced by the instantaneous compression scheme in terms of speech recognition are observed.

Variability in the clear speech intelligibility advantage

The overall intelligibility advantage for sentences produced in clear versus conversational speech is well-documented. This study looked at recognition accuracy across words in early and late positions in semantically anomalous and meaningful sentences spoken in clear versus conversational speaking styles. For both sentence types, the results showed the expected overall intelligibility advantage for clear speech over conversational speech. For the semantically anomalous sentences, in both speaking styles, a decline in keyword identification rate was observed with words earlier in the sentence being more accurately recognized than words later in the sentence. Furthermore, the intelligibility advantage for clear over conversational speech remained relatively constant across word positions. For the meaningful sentences, the decline in keyword identification rate across word positions was observed for conversational speech only. Meaningful sentences spoken in clear speech yielded a high, relatively stable word identification rate across position-in-sentence, resulting in a larger clear speech intelligibility benefit for words late in the sentence than for words early in the sentence. These results suggest that for typical meaningful sentences, the acoustic-phonetic enhancements of clear speech and the availability of semantic-contextual information combine to ‘‘boost’’ the intelligibility of words in late sentence positions.

Serial Audiometry in a Clinical Trial of AIED Treatment

We analyzed pure-tone and speech audiometric results from a prospective trial of anti-inflammatory treatment of subjects with active autoimmune inner ear disease (AIED). We sought to characterize the pattern and size of the treatment effect as reflected in clinical audiometry and to identify audiometric predictors of response to steroid treatment of AIED. Adult participants demonstrated clinically established criteria for AIED (n = 116). Eligibility required audiometric evidence of active AIED as indicated by idiopathic sensorineural hearing loss with threshold elevations within 3 months of enrollment. We evaluated audiometric changes after 4 weeks of treatment with pharmacologic doses (60 mg/day) of prednisone. We examined the relationship between audiometric pure-tone thresholds at baseline and changes in word intelligibility score (WIS) using parametric and nonparametric analyses. Magnitudes of change were assessed using independent or paired t-tests. Separate analyses were performed on subgroups that did or did not show improved WIS score with steroid treatment. Overall mean pure-tone averages improved from baseline to closeout of prednisone treatment in better hearing ears from 52.4 to 48.3 dB (p < .0001). Mean WIS improved in the better ear from 71.4% to 78.1% (p < .0001). Of pure-tone measures, only the six-tone average showed significant correlation with both the absolute improvements in WIS and with the percentage change in WIS after treatment. Individual frequencies at baseline showed no significant relationship with changes in WIS score after treatment. In 69 (59.5%) of 116 subjects, WIS improved (range, 2-80%) in the better ear. In these subjects, the baseline pure-tone thresholds and pure-tone averages correlated significantly and positively with improvement in WIS. Steroid treatment in AIED-mediated hearing loss produce variable but significant hearing gains. Neither a focal, cochleotopic region of greatest vulnerability to AIED nor frequency-specific amenability to treatment were evident. We did observe that analysis of predictors and the degree of treatment effect vary with different approaches to measuring change in the WIS. Depending on the approach adopted, the size of the treatment effect may be greatest across intermediate hearing levels at baseline. These observations offer an audiometric database that may enable greater precision in judging clinically meaningful parameters for future studies of AIED treatment and other interventions for sensorineural hearing loss.