Prosodic analysis of neutral, stress-modified and rhymed speech in patients with Parkinson's disease

https://doi.org/10.1016/j.cmpb.2015.12.011Get rights and content

Highlights

  • We analysed neutral, stress-modified and rhymed speech in Parkinson's disease.

  • We proposed quantitative prosodic analysis of poem recitation task.

  • We showed rhythmical demands improve identification of hypokinetic dysarthria.

  • We introduced a concept of permutation test in dysarthric speech analysis.

Abstract

Background and objective

Hypokinetic dysarthria (HD) is a frequent speech disorder associated with idiopathic Parkinson's disease (PD). It affects all dimensions of speech production. One of the most common features of HD is dysprosody that is characterized by alterations of rhythm and speech rate, flat speech melody, and impairment of speech intensity control. Dysprosody has a detrimental impact on speech naturalness and intelligibility.

Methods

This paper deals with quantitative prosodic analysis of neutral, stress-modified and rhymed speech in patients with PD. The analysis of prosody is based on quantification of monopitch, monoloudness, and speech rate abnormalities. Experimental dataset consists of 98 patients with PD and 51 healthy speakers. For the purpose of HD identification, sequential floating feature selection algorithm and random forests classifier is used. In this paper, we also introduce a concept of permutation test applied in the field of acoustic analysis of dysarthric speech.

Results

Prosodic features obtained from stress-modified reading task provided higher classification accuracies compared to the ones extracted from reading task with neutral emotion demonstrating the importance of stress in speech prosody. Features calculated from poem recitation task outperformed both reading tasks in the case of gender-undifferentiated analysis showing that rhythmical demands can in general lead to more precise identification of HD. Additionally, some gender-related patterns of dysprosody has been observed.

Conclusions

This paper confirms reduced variation of fundamental frequency in PD patients with HD. Interestingly, increased variability of speech intensity compared to healthy speakers has been detected. Regarding speech rate disturbances, our results does not report any particular pattern. We conclude further development of prosodic features quantifying the relationship between monopitch, monoloudness and speech rate disruptions in HD can have a great potential in future PD analysis.

Introduction

Parkinson's disease (PD) is the second most frequent neurodegenerative disorder [1] with the prevalence rate estimated to 1.5% for people aged over 65 years [2]. It was first described in 1817 by James Parkinson in his essay “An Essay on the Shaking Palsy” that provided an overview of the disease in its medical context. Parkinson himself referred to it as paralysis agitans and the term Parkinson's Disease was established later by Jean-Martin Charcot in 1876. PD is a chronic idiopathic disorder with the unknown aetiology that is characterized by the progressive loss of dopaminergic neurons in substancia nigra pars compacta [3].

Deficiency of dopamine is a major cause of parkinsonian symptoms [4]. It leads to malfunction of the central nervous system that can no longer co-ordinate muscle movements appropriately resulting in variety of associated motor and non-motor deficits that significantly reduce the quality of life of patients with PD. In addition to motor symptoms manifested predominantly on upper and lower extremities such as resting tremor, bradykinesia, muscular rigidity [3] and postural instability, patients with PD also develop behavioural alternations, reduction of cognitive abilities and the previous studies showed that one of the most frequent manifestations of PD is a multi-modal disruption of motor speech realization commonly referred to as hypokinetic dysarthria (HD) [5]. The latest studies of PD report that speech disorders are amongst the earliest indicators of PD onset. Furthermore it has been estimated that approximately 60–90% of patients with PD develop HD [6].

Generally, human speech can be briefly defined as the ability to create reproducible sounds with certain semantic meanings, produced by coordinated movement of more than 100 muscles [7], [8]. The ability to speak is a unique and complex process, which can be subdivided into several dimensions, namely: respiration, phonation, articulation, and prosody [9]. HD affects all dimensions of speech [10], [11]. The related speech disorders can be summarized as follows: voice tremor [3], hypophonia [12], dysphonia [13], dysprosody [14], increased voice nasality [15], articulation problems in combination with unorthodox speech silence and speech rate alterations such as hesitation [16], palilalia [16], bradyphemia [17] or tachyphemia [17]. Accordingly, negative impact of HD on daily life in PD has been emphasized by several researchers so far [18], [19].

At present, there is no definite cure for PD and the disease is ultimately fatal. Medical treatment of PD is consequently limited to reduction of its cardinal motor symptoms. Conventional treatment of PD consists of pharmacological medication in form of levodopa (L-dopa) [20], which is known to be efficient for treating certain parkinsonian manifestations such as bradykinesia, muscular rigidity and resting tremor on upper and lower extremities. However, the efficiency of L-dopa for alleviating symptoms of HD is insufficient [11]. Moreover, the current diagnosis of PD is based on the presence of primary motor symptoms, which are known to develop after 60–70% of dopaminergic neurons are degenerated [21], [22]. In its early stage, PD is very hard to recognize and the symptoms get worse with time. Considering the fact that the effect of PD treatment depends upon the stage of the disease during which it is initiated and the impossibility to halt PD progression, we conclude that the accurate PD rating is critical for effective PD treatment and reduction of its further development that can generally prolong active life expectancy and also mitigate the loss of communication and frequent social isolation of PD patients.

Taking into account all the previously mentioned information, there is an increasing demand for future accurate para-clinical monitoring of PD symptoms. The essential benefits of para-clinical methods in comparison with the clinical ones are the associated non-invasiveness, speed and inexpensiveness (speech can be self-recorded with no need of supervision or expensive equipment) and most importantly its natural objectivity [13], [23]. This paper focuses on one of the most commonly used paraclinical method of PD assessment, namely the acoustic analysis of dysarthric speech.

The prosodic features of human speech describe stress and rhythm, speech to pause ratio and velocity, speech intensity and pitch variation [9], [24]. Prosody is an important aspect of human verbal communication. It conveys semantic, syntactic and affective information and also reflects emotions of the speaker. Due to vocal tract muscle stiffness [25] patients with PD exhibit alterations of rhythm and speech rate (inappropriate silences, short rushes of speech, and variable speech rate) [14], imprecise articulation, small variations in pitch and intensity (monopitch and monoloudness) resulting into flat voice melody lacking intonation [26], [9]. Such deterioration of prosody has a detrimental impact on speech naturalness and intelligibility leading into substantial voice and speech quality deficits.

Intonation is defined as an ensemble of pitch variations during an utterance [27]. According to the previous studies, patients with PD indicated a reduction of fundamental frequency (F0) variability in comparison with healthy speakers [26], [28], [29], [30]. Following their previous research [8], Skodda et al. [31] investigated F0 variability and speech rate in patients with PD during a reading task. Altogether, they recruited 138 PD patients and 50 healthy controls (HC). Based on the acoustic analysis, PD patients showed significantly reduced variability of pitch compared to healthy speakers when related to the entire reading task, which is in accordance with the previous research [28], [29], [32] performed in this field of science. Moreover, a notable decline of pitch variation was observed over the course of the reading. In 2011, Rusz et al. [33] demonstrated that even in the early stages of PD, patients do exhibit lower melody intonation and decreased intensity variations referring to limited range of motions and impaired laryngeal tension in combination with insufficient breath support. Results of this study indicated that amongst all known manifestations of HD, patients in the early stage of PD suffer primarily from prosodic impairment. Moreover, Rusz and colleagues [33] reported that reduced melody variations in parkinsonian speech can be related to the lowered ability of stress pronunciation and emotional intonation limitation.

In 2014, Tykalova et al. [34] investigated contrastive stress in 20 male patients in early stage of PD and 16 age- and gender-matched HC. The participants were asked to unnaturally emphasize several key words during a reading task. In contrast with HC, patients with PD showed a distinctively flatter pitch contours, especially at the beginning and the end of a phrase, confirming the previous findings of monopitch intonation in parkinsonian dysprosody [33], [35]. In addition, Pell et al. [18] investigated impact of dysprosody on vocal-prosodic communication from the perspective of listeners. They reported that listeners experienced serious difficulties recognizing the emotional prosody (with the exception of anger), sentence mode, phonemic stress and contrastive stress of patients with PD in comparison with healthy speakers. Furthermore, previous studies have reported several cases in which monopitch intonation had been observed several years before the onset of cardinal motor symptoms [36]. Since the monopitch intonation is present in a majority of patients with PD along with its probability to occur even in the early stages of the disease [37], the accurate monitoring of F0 variability is necessary.

As reported by Duffy in [7], monoloudness is a common prosodic flaw in PD. Back in 1986, Metter and Hanson reported a study in which they showed that patients with PD produced a significantly smaller intensity variation compared to healthy speakers during the reading of a standard passage [28]. Later, according to Watson and Munson, PD speakers exhibit overall lower speech intensity, deficits in intensity range, and intensity variations during speech production [38]. In addition, Skodda et al. proposed a gender-related study of dysprosody in PD [39]. This study comprised 169 PD patients and 64 age-matched healthy speakers. They reported a reduction in speech intensity during a reading task composed of 4 sentences. In 2011, Rusz et al. [33] showed that early-staged PD patients do exhibit a decreased intensity variations in comparison with Clark et al. [40] performed a study in which they examined loudness perception in PD (17 PD patients and 25 HC) and reported that patients with PD produced a significantly different pattern and more restricted range compared to healthy speakers in their perception of speech loudness. These results demonstrate the importance of the analysis and treatment (e.g. Lee Silverman Voice Treatment, see [41], [42], [43], [44]) of speech intensity deterioration.

Regarding the analysis of speech rate disturbances in PD, back in 1963 Canter [45] found no differences in number of pauses or mean pause duration between patients with PD and healthy speakers during a reading task. However, several years later, Metter et al. [28] demonstrated the presence of these speech rate abnormalities. Amongst more recent studies, in 2008, Skodda and Schlegel [8] analysed articulatory rate and pause time during a standardized reading task (170-syllabic text composed of 4 complex sentences) in 121 PD patients and 70 healthy speakers. They investigated the performance of acoustic measurements applied on the first and the last sentence in order to evaluate the hypothesis of altered speech rate and rhythm in PD. Results of this study confirmed an age-related reduction of articulation rate, proposed by Weismer [46] in 1984.

Furthermore, a gender specific aspects of speech rhythm was referred. However, neither gender-related differences between PD patients and HC in speech rate parameters (total speech rate, net speech rate), nor overall distinctions in speech rate between PD and control subjects were observed. In 2009, Skodda et al. performed a longitudinal study [25] in which they found a speech rate variation (number of syllables per second) closely related to the progression of PD and confirmed the previous findings of Ho et al. [5], in which decisive impairment of speech fluency was found in the more severe stages of the disease. Next, in [31], Skodda et al. investigated net speech rate of patients with PD during a reading task. No significant distinctions of net speech rate between PD patients and HC were observed. In addition, increased number of pauses per second during a reading task were reported by Skodda et al. in [39].

According to the previous studies, a comprehensive research has been applied in this field of science along with the vast development of wide range of speech tests quantifying various aspects of parkinsonian dysarthria, namely: fast syllable repetition (diadocho-kinetic (DDK) tasks) [47], [48], [49], sentence repetition tasks [44], [50], numerous reading tasks [39], [51], or running-speech (monologue) tasks [25], [49], [52], [53], [54]. DDK tasks have been used several times to assess mainly the articulatory impairment primarily caused by the presence of HD [47], [48], [49]. However, several researchers used DDK tasks to investigate dysprosody in PD, e.g. [55]. Regarding sentence repetition task, Bandini et al. [50] have recently employed a study in which they evaluated patterns of dysprosody in patients with idiopathic PD (14 males and 6 females) via a fully automated tool. They showed that PD patients exhibit longer pauses between each sentence repetition confirming the presence of speech rate deficits in HD. According to the research proposed by Rusz et al. [49], [54] monologue is the most prospective speech task to assess the speech impairment in PD. However, as reported by Bandini and colleagues [50], the assessment of speech alternations from monologue requires a development of precise and automatic approach, which to our best knowledge has not been standardized yet.

Over the last few decades, researchers have performed studies in which they investigated many aspects of prosodic abnormalities caused by neurodegenerative disorders such as PD [39], [56], [57]. These studies comprised examination of intonation and melody of speech, variation in speech loudness and timing, contrastive stress and rhythmical deficits. Regarding speech intonation and intensity, Anand and Stepp [58] have recently demonstrated that pitch and intensity variation and the associated naturalness of speech perceived by listeners are strongly correlated. Concerning speech rate disturbances in PD, there are many papers describing the speech rate deficits in PD, confirming the fact that rigidity and hypokinesia of the laryngopharyngeal tractus [12] has a detrimental effect on speech rate and timing [25], [31], [39]. However, in 2008 Lowit [55] investigated rhythmical impairment in PD and suggested that it is not sufficient to only capture duration-based characteristics without considering their relationship with parameters such as: F0 and intensity of speech.

In summary, reduced variability of pitch, monoloudness and abnormalities in speech rate and pausing have been observed [8], [26], [28], [29], [30], [33], [35], [59]. Whereas monopitch intonation is proven in many articles to sufficiently characterize parkinsonian dysprosody, the results of the previous studies on articulatory velocity and speech rate in patients with PD remains inconsistent [24], [29], [60]. As reported by Skodda et al. [39] examination of speech rate in PD is task related (syllable repetition may rather indicate articulatory capacities, whereas speech rate characteristics calculated from sentences, standardized reading tasks, or a free monologue are likely to reflect prosodic aspects of HD instead). Several other researchers [61], [62], [63] have also pointed out the fact that basal ganglia processing demands differ based on the selection of the associated speech task [32], [64]. Furthermore, the results of the previous studies differ significantly in the sample sizes [24], [29], [60], [65] used to assess the speech impairment in HD. Therefore, subsequent research is warranted to offer a deeper understanding of manifestation of PD on the human speech production.

To sum up the introduction, in this paper we perform a complex quantitative analysis of dysprosody in PD. As summarized in the preceding paragraphs, many researchers used reading tasks to assess the prosodic impairment in HD [8], [28], [31], [34], [39], [45], [51]. Therefore to obtain a comparable results, we follow these papers and use reading tasks to investigate dysprosody in PD, namely: (a) reading a short paragraph with neutral emotion; (b) stress-modified reading task (consisted of interrogative, imperative and indicative sentences). Moreover, we introduce the analysis of a poem recitation task. Using this approach we aim to compare the reading task with neutral emotion with the stress-modified reading task and also compare the reading tasks (without rhythmical patterns) with the rhythmically modified recitation task, which is a novel approach in this field of science. To our best knowledge a quantitative analysis of poem recitation task has not been used for the purpose of describing dysprosody in PD patients before. As far as we know, the only related analysis of rhythm in PD was performed by Rusz et al. in [33], however they used a speech task in which the speakers rhythmically read text consisted of 8 rhymes. The aims of this paper can be summarized as follows:

  • 1.

    To compare the selected prosodic features expressing monopitch intonation (standard deviation of fundamental frequency), monoloudness (standard deviation of squared energy operator) and speech rate abnormalities (net speech rate) between PD patients and HC.

  • 2.

    Regarding the speech tasks, we aim to compare the performance of these tasks to assess the prosodic impairment in HD: (a) reading task with neutral emotion vs. stress-modified reading task; (b) the reading tasks vs. poem recitation task.

  • 3.

    Concerning the speech analysis scenarios, we aim to investigate the classification performance of the following features: (a) features describing monopitch intonation; (b) features expressing monoloudness; (c) features quantifying speech rate abnormalities; and (d) general prosodic features (composed of all three prosodic subsets).

Section snippets

Patients and data acquisition

A grand total of 149 Czech native speakers participated in this study. Altogether, 98 patients with idiopathic PD (59 men/39 women; mean age 67.52 ± 8.29 years; mean disease duration 7.80 ± 4.42 years; UPDRS III score 24.91 ± 11.97; RBDSQ score 3.79 ± 3.23; FOG-Q score 7.16 ± 5.81; BDI 13.19 ± 14.57; daily levodopa equivalent dose (LED) 1005.93 ± 545.66 mg), see Table 1 and 51 healthy controls (25 men/26 women; mean age 63.96 ± 9.21 years) were enrolled at the First Department of Neurology, St. Anne's

Results

We began the exploration of the data by computing the relevance of the prosodic features to the participants’ clinical diagnosis (PD/HC). The analysis of the prosodic features by Spearman's rank correlation, mutual information, Mann–Whitney U test and a single feature classification can be seen in Table 2. The table shows top three features strongly associated to the response for females, males and a mixture of both genders. As can be seen in Table 2 the best classification performance in terms

Conclusion

In this paper we performed a complex quantitative analysis of dysprosody in 98 patients with idiopathic PD and 51 HC. Demographic and clinical characteristics of the patients are summarized in Table 1. As mentioned in [39], a fundamental assumption to obtain meaningful results is the comparison of the data obtained from PD patients with the ones acquired from an age- and gender-matched group of control speakers. Therefore, with respect to the previous studies that reported a significant impact

Acknowledgements

Research described in this paper was financed by the National Sustainability Program under grant LO1401. For the research, infrastructure of the SIX Center was used.

References (89)

  • S. Skodda et al.

    Gender-related patterns of dysprosody in Parkinson's disease and correlation between speech variables and motor symptoms

    J. Voice

    (2011)
  • J.P. Clark et al.

    Loudness perception and speech intensity control in Parkinson's disease

    J. Commun. Disord.

    (2014)
  • F. Rudzicz

    Using articulatory likelihoods in the recognition of dysarthric speech

    Speech Commun.

    (2012)
  • S. Skodda et al.

    Vowel articulation in Parkinson's disease

    J. Voice

    (2011)
  • A. Bandini et al.

    Automatic identification of dysprosody in idiopathic Parkinson's disease

    Biomed. Signal Process. Control

    (2015)
  • D. Kempler et al.

    Effect of speech task on intelligibility in dysarthria: a case study of Parkinson's disease

    Brain Lang.

    (2002)
  • A.K. Ho et al.

    Sequence heterogeneity in parkinsonian speech

    Brain Lang.

    (1998)
  • C.L. Ludlow et al.

    Speech timing in Parkinson's and Huntington's disease

    Brain Lang.

    (1987)
  • N. Giladi et al.

    Construction of freezing of gait questionnaire for patients with parkinsonism

    Parkinsonism Relat. Disord.

    (2000)
  • M. Folstein et al.

    “Mini-mental state”: A practical method for grading the cognitive state of patients for the clinician

    J. Psychiatr. Res.

    (1975)
  • H. Deng et al.

    Gene selection with guided regularized random forest

    Pattern Recognit.

    (2013)
  • B.W. Matthews

    Comparison of the predicted and observed secondary structure of T4 phage lysozyme

    Biochim. Biophys. Acta (BBA)

    (1975)
  • E. Combrisson et al.

    Exceeding chance level by chance: the caveat of theoretical chance levels in brain signal classification and statistical assessment of decoding accuracy

    J. Neurosci. Methods

    (2015)
  • M.C. de Rijk et al.

    Prevalence of Parkinson's disease in Europe: a collaborative study of population-based cohorts

    Neurology

    (2000)
  • M.C. de Rijk et al.

    Prevalence of parkinsonism and Parkinson's disease in Europe: the Europarkinson collaborative study. European community concerted action on the epidemiology of Parkinson's disease

    J. Neurol. Neurosurg. Psychiatry

    (1997)
  • O. Hornykiewicz

    Biochemical aspects of Parkinson's disease

    Neurology

    (1998)
  • P. Brodal

    The Central Nervous System: Structure and Function

    (2003)
  • A.K. Ho et al.

    Speech impairment in a large sample of patients with Parkinson's disease

    J. Behav. Neurol.

    (1999)
  • L.O. Ramig et al.

    Speech treatment for Parkinson's disease

    Expert Rev. Neurother.

    (2008)
  • J.R. Duffy

    Motor Speech Disorders: Substrates, Differential Diagnosis, and Management

    (2013)
  • S. Skodda et al.

    Speech rate and rhythm in Parkinson's disease

    Mov. Disord.

    (2008)
  • F.L. Darley et al.

    Motor Speech Disorders

    (1975)
  • J. Mekyska et al.

    Selection of optimal parameters for automatic analysis of speech disorders in Parkinson's disease

  • I. Eliasova et al.

    Acoustic evaluation of short-term effects of repetitive transcranial magnetic stimulation on motor aspects of speech in Parkinson's disease

    J. Neural Transm.

    (2013)
  • K.K. Baker et al.

    Thyroarytenoid muscle activity associated with hypophonia in Parkinson's disease and aging

    Neurology

    (1998)
  • A. Tsanas et al.

    Nonlinear speech analysis algorithms mapped to a standard metric achieve clinically useful quantification of average Parkinson's disease symptom severity

    J. R. Soc. Interface

    (2010)
  • M.F. Brin et al.

    Movement Disorders of the Larynx

    (2009)
  • R. Moretti et al.

    Speech initiation hesitation following subthalamic nucleus stimulation in a patient with Parkinson's disease

    Eur. Neurol.

    (2003)
  • M. Gentil et al.

    Parkinsonian dysarthria

    Rev. Neurol.

    (1995)
  • D. Hall et al.

    Pragmatic communication is impaired in Parkinson disease

    Int. J. Neurosci.

    (2011)
  • J.Y. Lee et al.

    Daily dose of dopaminergic medications in Parkinson's disease: clinical correlates and a posteriori equation

    Neurol. Asia

    (2010)
  • A. Tsanas et al.

    Accurate telemonitoring of Parkinson's disease progression by noninvasive speech tests

    IEEE Trans. Bio-Med. Eng.

    (2010)
  • S. Skodda et al.

    Progression of dysprosody in Parkinson's disease over time-a longitudinal study

    Mov. Disord.

    (2009)
  • G.J. Canter

    Speech characteristics of patients with Parkinson's disease: I. Intensity, pitch, and duration

    J. Speech Hear. Disord.

    (1963)
  • Cited by (59)

    • A cross-linguistic perspective to classification of healthiness of speech in Parkinson's disease

      2022, Journal of Neurolinguistics
      Citation Excerpt :

      Deviant speech dimensions have been extensively studied on an acoustic level. There are many studies focusing on the acoustic aspect of speech production in PD with attention to monopitch, or in acoustic terms - fundamental frequency (f0) deviations (Galaz et al., 2016; Skodda et al., 2013), distorted rhythm of speech (Skodda & Schlegel, 2008), monoloudness - reduced intensity variability of voice (Galaz et al., 2016; Skodda et al., 2013), reduced stress (Tykalova et al., 2014), imprecise consonants (Fischer & Goberman, 2010; Tykalova et al., 2017) and a hoarse and breathy voice quality (Tsanas et al., 2009). A number of studies have also demonstrated that prosody deficits together with harsh voice and reduced articulation are among the most prominent speech characteristics that are present in acoustics of speech produced by PwPD (Brabenec et al., 2017; Galaz et al., 2016; Rusz et al., 2011; Verkhodanova et al., 2019).

    View all citing articles on Scopus
    View full text