Comparability of modern recording devices for speech analysis: smartphone, landline, laptop, and hard disc recorder.

BACKGROUND: Large-scale multi-site experimental and clinical speech protocols require high-fidelity, easy-to-use speech recording technologies. However, little is known about the reliability and comparability of affordable, portable and commonly used technologies with traditional well-validated devices (e.g., a hard disc recorder with a high-quality microphone). OBJECTIVE: To examine the comparability of speech and voice samples acquired from protocols involving high- and low-quality devices. METHODS: Speech samples were acquired simultaneously from 15 healthy adults using four devices and analyzed acoustically for measures of timing and voice quality. For the purpose of making initial comparisons, methods were deemed comparable if the resultant acoustic data yielded root mean squared error values ≤10% and statistically significant Spearman's correlation coefficients. RESULTS: The data suggest that there is significant and widespread variability in the quality and reliability between different acquisition methods for voice and speech recording. Not one method provided statistically similar data to the protocol using the benchmark device (i.e., a high-quality recorder coupled with a condenser microphone). Acoustic analysis cannot be assumed to be comparable if different recording methods are used to record speech. CONCLUSIONS: Findings have implications for researchers and clinicians hoping to make comparisons between labs or, where lower-quality devices are suggested, to offer equal fidelity.

Longitudinal change in dysarthria associated with Friedreich ataxia: a potential clinical endpoint.

CNS functions that show change across short periods of time are particularly useful clinical endpoints for Friedreich ataxia. This study determined whether there is measurable acoustical change in the dysarthria associated with Friedreich ataxia across yearly intervals. A total of 29 participants diagnosed with Friedreich ataxia were recorded across 4 years at yearly intervals. A repeated measures ANOVA was used to determine which acoustic measures differed across time, and pairwise t tests were used to assess the consistency of the change across the time intervals. The relationship between the identified measures with perceptual severity was assessed with stepwise regression. Significant longitudinal change was observed with four measures that relate to the utterance duration and spectral changes in utterances. The spectral measures consistently detected change across time intervals of two or more years. The four measures combined moderately predicted perceptual severity. Together, the results implicate longitudinal change in speaking rate and utterance duration. Changes in speech associated with Friedreich ataxia can be measured across intervals of 2 years and therefore show rich potential for monitoring disease progression and therapy outcomes.

Differentiating profiles of speech impairments in Friedreich's ataxia: a perceptual and instrumental approach.

BACKGROUND: The speech disorder associated with Friedreich's ataxia (FRDA) is classically described as ataxic dysarthria. However, variable neuropathology beyond the cerebellum, which may include the corticospinal and corticobulbar tracts, means that the dysarthria can be mixed rather than a pure ataxic dysarthria. AIMS: To characterize physiological features of the dysarthria associated with FRDA and identify differential patterns of deviation that may occur across the subsystems of the speech-production mechanism in a series of seven case studies. METHODS & PROCEDURES: The assessment battery included a perceptual analysis of a speech sample using an interval rating scale, and a range of instrumental measures to investigate the respiratory, laryngeal, velopharyngeal and articulatory systems. OUTCOMES & RESULTS: The results demonstrated the variability that exists in the dysarthria associated with FRDA, highlighting the existence of differential profiles of speech impairment. A particular distinction was observed between the presence of hypernasality and phonatory dysfunction, as evidenced by the instrumental results. CONCLUSIONS & IMPLICATIONS: The distinct profiles of dysarthria associated with FRDA indicate that approaches that address multiple subsystems are necessary for the accurate characterization and quantification of the motor speech disorder. Further research is required to investigate the decline in speech function as the disease progresses, as changes in speech function over time may be a good indicator of neurological decline in FRDA.

Articulatory kinematics in the dysarthria associated with Friedreich's ataxia.

Electromagnetic articulography (EMA) was used to investigate the tongue kinematics in the dysarthria associated with Friedreich's ataxia (FRDA). The subject group consisted of four individuals diagnosed with FRDA. Five nonneurologically impaired individuals, matched for age and gender, served as controls. Each participant was assessed using the AG-200 EMA system during six repetitions of the tongue tip sentence Tess told Dan to stay fit and the tongue back sentence Karl got a croaking frog. Results revealed reduced speed measures (i.e., maximum acceleration / deceleration / velocity), greater movement durations and increased articulatory distances for the approach phases of consonant productions. The approach phase, involving movement up to the palate, was more affected than the release phase. It is suggested that deviant lingual kinematics could be the outcome of disturbances to cerebellar function, or possibly in combination with disturbances to upper motor neuron systems.

Spectral measures of the effects of Friedreich's ataxia on speech.

This study identifies two measures of the effects of Friedreich's ataxia (FRDA) on speech motor control. Speech samples of 17 healthy controls and 37 speakers with dysarthria associated with FRDA were recorded during one structured and one unstructured speaking task. Two measures of spectral variation were used that relate to the rate and range of changes that occur in the spectral envelope. Linear mixed models revealed significant effects of GROUP, TASK, and GROUP*TASK. FRDA speech samples had slower rate of spectral change and reduced spectral range. Healthy speakers produced faster rates of spectral change in read text compared to conversation, but speakers with dysarthria did not. The results suggest that structured speaking tasks which demand large spectral variation may be particularly useful in assessing the dysarthria. It is concluded that the rate of spectral change is a useful measure of dysarthria associated with FRDA.

Measures of spectral change and their application to habitual, slow, and clear speaking modes.

Spectral measures are sensitive to dysarthric speech. However, it is unclear whether the spectral differences in dysarthric and healthy speech are due to slow articulation rate or reflect other qualitative changes in speech. Spectral measures were used to detect differences between habitual, slow, and "clear" speaking modes in 12 healthy speakers. Matched t-tests were used to determine differences in the rate and degree of spectral change between the speaking modes. Pearson's correlation coefficients were calculated to assess how well rate of spectral change predicts articulation rate (syllables per second). Clear speech had a significantly higher degree of spectral change than habitual speech, and slow speech had a significantly slower rate of spectral change than habitual and clear speaking modes. These differences occurred in all 12 speakers. The rate of spectral change was correlated with articulation rate across all speakers (range of r = .8-.9 within individual speaking modes) and therefore is a gross predictor of articulation rate. These results suggest that measures of the degree and rate of spectral change together can be used to detect changes between clear, slow, and habitual speaking modes, and hold potential as performance measures.

Differentiating impairment levels in temporal versus spatial aspects of linguopalatal contacts in Friedreich's ataxia.

Electropalatography (EPG) was used to describe the pattern of linguopalatal contact and the consonant phase durations exhibited by a group of seven individuals with dysarthria associated with Friedreich's ataxia (FRDA). A group of 14 non-neurologically impaired individuals served as controls. The Reading Electropalatograph (EPG3) system was used to record linguopalatal contact during production of the target consonants (/t/, /l/, /s/, /k/) elicited in five words of CV and CVC construction, with the target consonants in word initial position. These words were embedded into short sentences and repeated five times by each participant. The FRDA group exhibited significantly increased consonant durations compared with the controls while maintaining normal linguopalatal contact patterns. These findings suggest that the articulatory impairment in FRDA manifests as a temporal rather than spatial disturbance.

Examining the effects of multiple sclerosis on speech production: does phonetic structure matter?

UNLABELLED: The second formant (F2) is well-known to be important to intelligibility (e.g. [Delattre, P., Liberman, A., & Cooper, F. (1955). Acoustic loci and transitional cues for consonants. Journal of the Acoustical Society of America, 27, 769-774]) and is affected by a variety of dysarthrias [Weismer, G., & Martin, R. (1992). Acoustic and perceptual approaches to the study of intelligibility. In R. Kent (Ed.), Intelligibility in speech disorders (pp. 67-118). Amsterdam: John Benjamins Publishing Company]. This study tests two related hypotheses: (1) dysarthria associated with Multiple Sclerosis (MS) has a greater effect on the ability to produce extreme F2 movement than on typical F2 movement and (2) phonetic stimuli associated with large and/or rapid F2 movement in healthy speakers precipitate larger differences between healthy and dysarthric speech than do stimuli associated with small and/or slow F2 movement. Twelve participants with MS and 16 healthy controls read aloud the Grandfather Passage. F2 slopes were calculated from the F2 tracings (i.e. change in Hz over a 20ms lag). For each sentence, the following measures of F2 movement were calculated: F2 range (maximum F2-minimum F2), median slope, and 95%ile slope. The mean and the maximum observation for each participant were respectively used as measures of typical and extreme productions. A repeated-measures MANOVA detected significantly larger group differences in the slopes of maximum productions than for mean production. Also, group differences in F2 slope were greater in phonetic stimuli associated with the largest F2 slopes in healthy speech than in phonetic stimuli associated with more shallow slopes. No group differences in F2 range were detected. Results indicate dysarthria affects the production of extremely rapid changes in F2 and that some phonetic structures are more useful than others for detecting these impairments. It is concluded that comparison of speakers' best-productions (e.g. maximum F2 slope) yield more useful estimates of the effects of mild to moderate dysarthria on F2 slopes than do typical productions. LEARNING OUTCOMES: The reader will be able to: (1) describe two approaches to measuring typical and extreme acoustic variability that can be applied to connected speech, (2) discuss evidence that MS affects extremely rapid changes in F2.

Parametric quantitative acoustic analysis of conversation produced by speakers with dysarthria and healthy speakers.

PURPOSE: This study's main purpose was to (a) identify acoustic signatures of hypokinetic dysarthria (HKD) that are robust to phonetic variation in conversational speech and (b) determine specific characteristics of the variability associated with HKD. METHOD: Twenty healthy control (HC) participants and 20 participants with HKD associated with idiopathic Parkinson's disease (PD) repeated 3 isolated sentences (controlled phonetic content) and 2 min of conversational speech (phonetic content treated as a random variable). A MATLAB-based program automatically calculated measures of contrastivity: speech-pause ratio, intensity variation, median and maximum formant slope, formant range, change in the upper and lower spectral envelope, and range of the spectral envelope. t tests were used to identify which measures were sensitive to HKD and which measures differed by task. Discriminant analysis was used to identify the combination of measures that best predicted HKD, and this analysis was then used as a general measure of contrastivity (Contrastivity Index). Differential effects of HKD on maximum and typical contrastivity levels were tested with interaction of maximum, minimum, and median observations of individual speakers and with pairwise comparisons of skewness and kurtosis of the contrastivity index distributions. RESULTS: Group differences were detected with pairwise comparisons with t tests in 8 of the 9 measures. Percentage pause time and spectral range were identified as the most specific (95%) and accurate (95%) differentiators of HKD and HC conversational speech. Sentence repetition elicited significantly higher levels of contrastivity than conversational speech in both HC and HKD speakers. Maximum and minimum contrastivities were significantly lower in HKD speech, but there was no evidence that HKD affects maximum contrastivity levels more than median contrastivity levels. The HKD speakers' contrastivity distributions were significantly more skewed to lower levels of production. CONCLUSION: HKD can be consistently distinguished from HC speech in both sentence repetition and conversational speech on the basis of intensity variation and spectral range. Although speakers with HKD were effectively able to produce higher contrastivity levels in sentence repetition tasks, they habitually performed closer to the lower end of their production ranges.

Task-based profile of vocal intensity decline in Parkinson's disease.

This study examines intensity decay in the phonation of persons with Parkinson's disease (PD). The decline in vocal intensity (determined by linear regression of the intensity envelope) was compared across the following tasks: vowel prolongation, syllable repetition (diadochokinesis, DDK), isolated sentences and conversation. In contrast to previous studies, PD speakers demonstrated no significant differences in intensity decline from healthy speakers in vowel prolongation. The vocal intensity of speakers with PD declined more rapidly than that of controls in DDK tasks. While intensity slopes in conversation were more variable in both groups, some participants with PD exhibited abrupt changes in intensity. Results indicate that the DDK is particularly useful for describing intensity decay associated with PD. However, considering the inconsistent group differences across tasks, and the discrepant findings from previous studies, intensity decay is not a robust symptom of PD.

Lognormal distribution of pause length in ataxic dysarthria.

Previous investigations indicated that pauses in ataxic dysarthric speech are abnormal, but little is known about the nature of their variation or whether it can be characterized by a common distribution. The following types of pauses were measured in spontaneous monologues of 16 speakers with ataxic dysarthria: pauses that contained evidence of respiration, pauses that occurred between words, between phonemes, within phonemes, and those associated with stop consonants. Results showed that the duration of pauses not associated with stop consonants can be modeled with two distinct lognormal distributions. The lognormal distribution predicted duration of across types within 1.67% standard error, with the exception of pauses that occurred within phonemes. Distributional parameters pointed to some fundamental similarities between pauses that occur within words. The robustness of the lognormal distribution across pause types strongly supports this distribution as a viable and useful method of analysis. We hope this work will encourage others to consider the lognormal distribution.