The Effect of Speaking Sample Duration on Determination of Habitual Pitch

Depressive and mania mood state detection through voice as a biomarker using machine learning

Introduction

Depressive and manic states contribute significantly to the global social burden, but objective detection tools are still lacking. This study investigates the feasibility of utilizing voice as a biomarker to detect these mood states. Methods:From real-world emotional journal voice recordings, 22 features were retrieved in this study, 21 of which showed significant differences among mood states. Additionally, we applied leave-one-subject-out strategy to train and validate four classification models: Chinese-speech-pretrain-GRU, Gate Recurrent Unit (GRU), Bi-directional Long Short-Term Memory (BiLSTM), and Linear Discriminant Analysis (LDA).

Results

Our results indicated that the Chinese-speech-pretrain-GRU model performed the best, achieving sensitivities of 77.5% and 54.8% and specificities of 86.1% and 90.3% for detecting depressive and manic states, respectively, with an overall accuracy of 80.2%.

Discussion

These findings show that machine learning can reliably differentiate between depressive and manic mood states via voice analysis, allowing for a more objective and precise approach to mood disorder assessment.

Quantitative Speech Assessment in Ataxia-Consensus Recommendations by the Ataxia Global Initiative Working Group on Digital-Motor Markers

Dysarthria is a common and debilitating symptom of many neurodegenerative diseases, including those resulting in ataxia. Changes to speech lead to significant reductions in quality of life, impacting the speaker in most daily activities. Recognition of its importance as an objective outcome measure in clinical trials for ataxia is growing. Its viability as an endpoint across the disease spectrum (i.e. pre-symptomatic onwards) means that trials can recruit ambulant individuals and later-stage individuals who are often excluded because of difficulty completing lower limb tasks. Here we discuss the key considerations for speech testing in clinical trials including hardware selection, suitability of tasks and their role in protocols for trials and propose a core set of tasks for speech testing in clinical trials. Test batteries could include forms suitable for remote short, sensitive and easy to use, with norms available in several languages. The use of artificial intelligence also could improve accuracy and automaticity of analytical pipelines in clinic and trials.

Analysis of Speech Fundamental Frequencies for Different Tasks in Japanese

PURPOSE

Speech fundamental frequency (SFF) assessment is essential for all dysphonia patients to effectively evaluate the therapeutic effects of voice therapy, especially in patients with disturbances in their voice pitch due to mutational dysphonia, Reinke's edema, or as side effects of hormone therapy. A standard method of SFF measurement remains unknown. Speech tasks such as sustained vowel phonation, counting, reading passage, and spontaneous speech have generally been used for SFF measurements. Ideally, spontaneous speech best reflects SFF; however, this task has not yet been clearly defined and is limited with regard to its adaptation to a clinical setting. A reliable task for SFF measurement in Japanese, which corresponds to a speech task that most closely reflects the value that would be observed with typical spontaneous speech, has not been investigated. This study aimed to identify a reliable speech task by measuring the SFF values elicited by different widely used speech tasks in Japanese, and assess its reliability and coefficient of determination (R²).

METHODS

Sixty healthy volunteers (30 men and 30 women; aged 19-30 years; mean age 22.5 years) were enrolled. All experimental procedures were performed in Japanese. The SFF values for the speech tasks were determined through the voice samples recorded using a Pulse Code Modulation (PCM) recorder. Each task, except spontaneous speech, was repeated five times, and the average fundamental frequency in each task was determined as the SFF. To assess the reliability of the SFF values across daily variations within individual speakers, the SFF measurements were repeated on two different days, separated by at least 1 week.

RESULTS

The SFF values of sustained /a/ phonation, sustained vowel-average, counting, reading passage, and spontaneous speech had excellent reliability, in terms of their reproduction based on intraclass correlation. Significantly high SFF values were observed, in decreasing order, for sustained vowels-average, counting, reading passage, and spontaneous speech in both males and females. The highest R² for spontaneous speech was that of reading passage in both males (R² = 0.771) and females (R² = 0.806) (P < 0.01).

CONCLUSION

When spontaneous speech was presented as a task most reflective of daily conversation, reading passage was determined to be the reliable task to assess the therapeutic effect of voice therapy in Japanese.

The reproducibility of bio-acoustic features is associated with sample duration, speech task and gender

Bio-acoustic properties of speech show evolving value in analyzing psychiatric illnesses. Obtaining a sufficient speech sample length to quantify these properties is essential, but the impact of sample duration on the stability of bio-acoustic features has not been systematically explored. We aimed to evaluate bio-acoustic features' reproducibility against changes in speech durations and tasks. We extracted source, spectral, formant, and prosodic features in 185 English-speaking adults (98 w, 87 m) for reading-a-story and counting tasks. We compared features at 25% of the total sample duration of the reading task to those obtained from non-overlapping randomly selected sub-samples shortened to 75%, 50%, and 25% of total duration using intraclass correlation coefficients. We also compared the features extracted from entire recordings to those measured at 25% of the duration and features obtained from 50% of the duration. Further, we compared features extracted from reading-a-story to counting tasks. Our results show that the number of reproducible features (out of 125) decreased stepwise with duration reduction. Spectral shape, pitch, and formants reached excellent reproducibility. Mel-frequency cepstral coefficients (MFCCs), loudness, and zero-crossing rate achieved excellent reproducibility only at a longer duration. Reproducibility of source, MFCC derivatives, and voicing probability (VP) was poor. Significant gender differences existed in jitter, MFCC first-derivative, spectral skewness, pitch, VP, and formants. Around 97% of features in both genders were not reproducible across speech tasks, in part due to the short counting task duration. In conclusion, bio-acoustic features are less reproducible in shorter samples and are affected by gender.

The Impact of Languages and Cultural Backgrounds on Voice Quality Analyses

OBJECTIVE

The aim of this study was to verify the influence of different cultural backgrounds and languages on the perception of voice quality. Another aim was to evaluate the correlation between clinical auditory-perceptual judgments with acoustic measurements, such as the acoustic voice quality index (AVQI) and the acoustic breathiness index (ABI), i.e., their effect on the concurrent validity.

METHODS

Brazilian and European participants rated 2 culturally and linguistically different datasets (i.e., 150 Brazilian Portuguese and 218 German voice sample sets), with each dataset containing concatenated voice samples of continuous speech and sustained phonation. The raters evaluated the vocal quality in terms of the overall voice quality and breathiness. Intra- and interrater reliabilities were also tested. The psychometric qualities of the AVQI and the ABI (concurrent validity and diagnostic accuracy) were analyzed, considering the perceptual ratings of Brazilian and European participants for each dataset.

RESULTS

Rater reliability decreased for nonnative participants when evaluating the nonnative dataset. Brazilian raters perceived the German voices as more severe, whereas German raters considered the overall voice quality of Brazilian voices as less severe than native raters. The concurrent validity of the AVQI and the ABI, i.e., the agreement with the auditory-perceptual evaluations, was high irrespectively of the group of participants. The diagnostic accuracy, which refers to the ability to identify a deviated voice, was higher, considering the perceptual evaluations of native listeners.

CONCLUSIONS

Brazilian raters classified voice quality as more deviated, and the Brazilian voice samples were measured to be less severe (a possible language characteristic). Further studies are necessary to verify whether there was a task or sample consequence or if adjustments to the AVQI and ABI equations are required for Brazilian Portuguese.

Effect of reading passage length on quantitative acoustic speech assessment in Czech-speaking individuals with Parkinson's disease treated with subthalamic nucleus deep brain stimulation

Little is known about the minimum sample length required for the stable acoustic assessment of speech in Parkinson's disease (PD). This study aimed to investigate the effect of the duration of the reading passage on the determination of reliable acoustic patterns in individuals with PD treated with subthalamic nucleus deep brain stimulation. A phonetically balanced reading text of 313 words was collected from 32 Czech persons with PD, and 32 age- and sex-matched healthy controls. The reading passage was segmented to produce ten sub-texts of increasing length ranging from a one- to a ten-segment-long sub-text. An error rate analysis was used to estimate the required stabilization value by evaluating the differences between the sub-texts and the entire text across seven hypokinetic dysarthria features. The minimum length of a reading passage equal to 128 words was found to be necessary for acoustic assessment, with similar lengths being required for the controls (120 words) and the two PD subgroups, including Parkinsonian individuals with a mild (126 words) and moderate (128 words) dysarthria severity. The current study provides important guidelines for the necessary sample length for future expert instrumental dysarthria assessments and assists in decreasing the time required for clinical speech evaluations.

Influence of the Voice Sample Length in Perceptual and Acoustic Voice Quality Analysis

OBJECTIVE

To analyze the variations that different voice sample length (VSL) has on the perceived degree of voice quality deviation and on the Acoustic Voice Quality Index (AVQI) accuracy.

METHODS

Voices of 71 subjects (53 dysphonic; 18 vocally health) were recorded: numbers 1-20 (42 syllables) + vowel/a/. Three different VSL were edited: VSL_long, 1-20 + 3 seconds vowel/a/; VSL_cust, customized length, were voiced-segments of the continuous speech had the same length of the vowel (mean = 18.73 syllables corresponding to 3 seconds of only-voiced segments) + 3 seconds vowel/a/; VSL_short, 1-10 (15 syllables) + 3 seconds vowel/a/. Three voice specialists perceptually judged the overall voice quality (G); 3 sessions were performed to evaluate each VSL variant. AVQI's precision and Spearman correlation were assessed.

RESULTS

The intra-rater reliability was "almost perfect" (kappa >0.826) for all evaluators in VSL_short; "substantial" (0.684) and "almost perfect" (0.897) in VSL_cust and "fair" (0.447) to "almost perfect" (1.000) in VSL_long. The inter-rater reliability was "moderate" (0.554) for VSL_long, "substantial" (0.622 and 0.618) for VSL_cust and VSL_short. The Gmean and AVQI_mean were perceived as more severe for longer samples and less severe for shorter samples. Considering the AVQI, VSL_short (r = 0.665) presented the higher correlation. VSL_cust presented the best area under the ROC curve (0.821). VSL_long and VSL_cust specificity was 100%, VSL_short specificity was 75%; higher sensitivity was observed for VSL_short (74%).

CONCLUSION

The voice quality outcomes changes for different VSLs. Longer VSLs seem to be perceived as more deviated, shorter VSLs seem to be more reliable and have better correlation with the acoustic analysis. The AVQI best accuracy was found at a customized length. Thus, to increase the voice analysis reliability, standardized procedure must be followed, including a precise speech material control allowing comparison among clinics and voice-centers.

Recommended Protocols for Instrumental Assessment of Voice: American Speech-Language-Hearing Association Expert Panel to Develop a Protocol for Instrumental Assessment of Vocal Function

PURPOSE

The aim of this study was to recommend protocols for instrumental assessment of voice production in the areas of laryngeal endoscopic imaging, acoustic analyses, and aerodynamic procedures, which will (a) improve the evidence for voice assessment measures, (b) enable valid comparisons of assessment results within and across clients and facilities, and (c) facilitate the evaluation of treatment efficacy.

METHOD

Existing evidence was combined with expert consensus in areas with a lack of evidence. In addition, a survey of clinicians and a peer review of an initial version of the protocol via VoiceServe and the American Speech-Language-Hearing Association's Special Interest Group 3 (Voice and Voice Disorders) Community were used to create the recommendations for the final protocols.

RESULTS

The protocols include recommendations regarding technical specifications for data acquisition, voice and speech tasks, analysis methods, and reporting of results for instrumental evaluation of voice production in the areas of laryngeal endoscopic imaging, acoustics, and aerodynamics.

CONCLUSION

The recommended protocols for instrumental assessment of voice using laryngeal endoscopic imaging, acoustic, and aerodynamic methods will enable clinicians and researchers to collect a uniform set of valid and reliable measures that can be compared across assessments, clients, and facilities.

Effects of age on the amplitude, frequency and perceived quality of voice

The manner and extent to which voice amplitude and frequency control mechanisms change with age is not well understood. The related question of whether the assessment of one's own voice evolves with age, concomitant with the acoustical changes that the voice undergoes, also remains unanswered. In the present study, we characterized the aging of voice production mechanisms (amplitude, frequency), compared the aging voice in different experimental contexts (vowel utterance, connected speech) and examined the relationship between voice self-assessment and age-related voice acoustical changes. Eighty healthy adults (20 to 75 years old) participated in the study, which involved computation of several acoustical measures of voice (including measures of fundamental frequency, voice amplitude, and stability) as well as self-assessments of voice. Because depression is frequent in older adults, depression and anxiety scores were also measured. As was expected, analyses revealed age effects on most acoustical measures. However, there was no interaction between age and the ability to produce high/low voice amplitude/frequency, suggesting that voice amplitude and frequency control mechanisms are preserved in aging. Multiple mediation analyses demonstrated that the relationship between age and voice self-assessment was moderated by depression and anxiety scores. Taken together, these results reveal that while voice production undergoes important changes throughout aging, the ability to increase/decrease the amplitude and frequency of voice are preserved, at least within the age range studied, and that depression and anxiety scores have a stronger impact on perceived voice quality than acoustical changes themselves.

The use of analogy in speech motor performance

The acoustic correlates of pitch variation were examined in 40 participants who received analogy instructions or explicit instructions that required them to modulate their intonation during speech production. First, using focus group methodology, professional speech-language pathologists were asked to identify analogies that best described minimum pitch variation (monotone), moderate pitch variation (normal intonation), and maximum pitch variation (exaggerated intonation) in speech. The focus group established that an appropriate pitch variation metaphor may be related to imagery of "waves at sea", with minimum pitch variation represented by a flat calm sea, moderate pitch variation represented by a moderate sea, and maximum pitch variation represented by a choppy sea. Forty adult participants without speech impairments were asked to read aloud a standard paragraph using their habitual pitch variation (control condition). They were then allocated randomly to an analogy or an explicit instruction group and were asked to read aloud different paragraphs with minimum, moderate, or maximum pitch variations. Results revealed that acoustic correlates of pitch variation (standard deviation of fundamental frequency, SDF0) were not different for the control condition, or moderate and maximum pitch variation conditions in the two groups. However, the analogy instruction was significantly more effective than the explicit instruction for inducing minimum pitch variation. Analysis of participants in each group who showed higher than normal pitch variation in the control condition (>.5 SD above the group SDF0) revealed that the analogy instruction was more effective than the explicit instruction in the minimum variation condition. It was concluded that analogy instructions may be a useful tool in speech rehabilitation.

Investigation of habitual pitch during free play activities for preschool-aged children

OBJECTIVE

This study is designed to compare the habitual pitch measured in two different speech activities (free play activity and traditionally used structured speech activity) for normally developing preschool-aged children to explore to what extent preschoolers vary their vocal pitch among different speech environments.

METHODS

Habitual pitch measurements were conducted for 10 normally developing children (2 boys, 8 girls) between the ages of 31 months and 71 months during two different activities: (1) free play; and (2) structured speech. Speech samples were recorded using a throat microphone connected with a wireless transmitter in both activities. The habitual pitch (in Hz) was measured for all collected speech samples by using voice analysis software (Real-Time Pitch).

RESULTS

Significantly higher habitual pitch is found during free play in contrast to structured speech activities. In addition, there is no showing of significant difference of habitual pitch elicited across a variety of structured speech activities.

CONCLUSIONS

Findings suggest that the vocal usage of preschoolers appears to be more effortful during free play than during structured activities. It is recommended that a comprehensive evaluation for young children's voice needs to be based on the speech/voice samples collected from both free play and structured activities.

The Effect of Speaking Context on Elicitation of Habitual Pitch

The purpose of this study was to investigate if there was an effect of speaking context on the elicitation of habitual pitch [speaking fundamental frequency (SFF)]. Six simulated speaking contexts were created (speaking during a voice evaluation, speaking in public, speaking to a peer, speaking to a superior, speaking to a subordinate, and speaking to a parent or spouse), and the SFF for 30 adult women with normal voice was compared across these contexts. A one-way analysis of variance (ANOVA) revealed a statistically significant (P < 0.001) effect of simulated speaking context on SFF, with post hoc analyses indicating a statistically significant difference in SFF while "speaking to a superior" (P < 0.001) and "speaking to a subordinate" (P < 0.001). Possible reasons for an effect of speaking context are discussed. Also, the implications of the use of varied speaking contexts when eliciting SFF are discussed, as is the possibility of an effect of speaking context on the elicitation of other clinically useful voice parameters.

The Effect of Speaking Task on Perceptual Judgment of the Severity of Dysphonic Voice

This study investigated the effect of speaking task on auditory-perceptual judgment of the severity of dysphonia. Three speech-language pathologists experienced in evaluating of disordered voices rated 29 recorded speakers, each of whom produced speech elicited via the same three tasks: sustained vowel /a/, oral reading of a standard passage, and connected speech describing a standard picture. Stimuli were played in sound field, and raters used direct magnitude estimation with a visual analog scale. A repeated measures analysis of variance (ANOVA) revealed a statistically significant (P < .05) effect of speaking task, with post hoc analyses that indicate a statistically significant difference between ratings for the sustained vowel versus connected speech elicited via picture description (P < .05). No statistically significant difference in ratings was found between oral reading and picture description or between oral reading and the sustained vowel. The ANOVA also revealed a statistically significant difference among raters (P < 0.001), but no statistically significant task by rater interaction. Possible reasons for an effect of speaking task are discussed. Also, the implications of one task for determination of dysphonic severity are discussed, as is the possibility of a task effect when determining other clinically useful vocal parameters.