The Comparison of Cepstral Peak Prominence by Using Speech Tool and Praat for Voice Analysis of Patients With Vocal Cord Palsy

Article information

J Korean Soc Laryngol Phoniatr Logop. 2023;34(2):45-49

Publication date (electronic) : 2023 August 29

doi : https://doi.org/10.22469/jkslp.2023.34.2.45

Jisu Kim, Min Young Cho, Dong Young Kim, Woongsang Sunwoo, Woori Park, Joo Hyun Woo

Department of Otorhinolaryngology-Head and Neck Surgery, Gil Medical Center, Gachon University College of Medicine, Incheon, Korea

Corresponding Author Woori Park, MD Department of Otorhinolaryngology-Head and Neck Surgery, Gil Medical Center, Gachon University College of Medicine, 21 Namdong-daero 774beon-gil, Namdong-gu, Incheon 21565, Korea Tel +82-32-460-3324 Fax+82-32-467-9044 E-mail wooripark@gilhospital.com

Received 2023 May 29; Revised 2023 July 26; Accepted 2023 August 4.

Abstract

INTRODUCTION

Patients with vocal cord palsy experience multiple difficulties, one of which is voice problems: breathy, weak, and low-pitched dysphonia.

Over the years of research on voice and speech, cepstral peak prominence (CPP) has become an objective measure of breathiness and dysphonia. Guidance from the American Speech-Language-Hearing Association recommends CPP as a tool for measuring the overall level of noise in the vocal signal and as a general measure of dysphonia [1].

Voice problems can be measured and quantified using different software programs, such as Analysis of Dysphonia in Speech and Voice, speech tools, and Praat [2].

As described by Maryn et al. [3], CPP is an acoustic measure of voice quality that has been qualified as the most promising and perhaps robust acoustic measure of dysphonia severity. Speech tool’s CPP value has been used in many studies to evaluate patient’s voice [4-8].

Different CPP measurement software programs provide different CPP values. In addition, software programs differ in terms of cost. Praat is a program that is free and open to society, which features different types of speech analysis including cepstral, pitch, formant, jitter, shimmer and many more. Praat can be a beneficial tool for the local outpatient clinic to use as an indicator of whether the patient’s voice has improved after a certain treatment.

Therefore, the purpose of this study is to compared the CPP values of speech tools and Praat voice analysis in patients with vocal cord palsy to determine whether there is a correlation between the CPP values of the two software programs and whether Praat’s CPP value could be beneficial in measuring the degree of voice problems at local outpatient clinics.

METHODS

RESULTS

Sixty patients were enrolled in the study. The vocal cord palsy group was older than the control group (Table 1). There were various reasons for vocal cord palsy, the most common of which was postoperative complications of thyroid surgery being 17 paitents (56.7%). Other causese of the vocal cord palsy included, post-concurrent chemoradiation therapy complication, cerebral infarct, viral infection and others were unknown. All the patients with vocal cord palsy were treated to alleviate their voice difficulties. Most of the patients underwent surgical procedures; 15 patients (50%) were treated with injection laryngoplasty, 13 patients (43.3%) were treated with injection laryngoplasty and arytenoid adduction, one patient was treated with laser cordectomy with medial arytenoidectomy and one patient was treated with voice therapy.

Table 1.

Subject characteristics

Speech tool vowels (STV) and speech tool sentences (STS) were compared with Praat vowels (PV) and sentences (PS) in the control, pre-treatment, and post-treatment groups (Table 2, Supplementary Figs. 1, 2 and 3 [in the online-only Data Supplement]). The STV and STS values were higher in the control group than in the pre-treatment vocal cord palsy group. After treatment, the patients showed better STV and STS scores. Similar results were observed in the PV and PS groups, with the highest scores observed in the control group and better scores after treatment in the vocal cord palsy group.

Table 2.

Mean cepstral peak prominence values of control, pretreatment, post-treatment group (p<0.001)

Vowel tests of both software programs showed more differences between the three groups than the sentence test (Fig. 1).

Fig. 1.

Mean cepstral peak prominence values of control, PTG and PoTG. CG, control group; STV, speech tool vowel; PV, Praat vowel; STS, speech tool sentence; PS, Praat sentence; PTG, pre-treatment group; PoTG, post-treatment group.

Most patients with vocal cord palsy undergo surgical procedures to improve their voice. Injection laryngoplasty was performed in almost all the patients. Patients who had unsatisfactory outcome after injection laryngoplasty underwent additional surgery such as arytenoid adduction. Patients whom underwent arytenoid adduction was chosen not based on Praat CPP scores but by their clinical evaluation and patients’ needs. The result showed lower ST and Praat values in the pre-treatment group than those in the injection laryngoplasty-only group. Better voice status was observed after treatment with ST and Praat. The PV presented the lowest value in the pre-treatment of patients with injection laryngoplasty and arytenoid adduction. Both software programs were statistically significant in representing voice improvement (Table 3).

Table 3.

Pre- and post-treatment value of ST and Praat in vocal cord palsy patients

DISCUSSION

Vocal cord paralysis causes serious phonation, respiration, and psychological problems that degrade quality of life. Vocal cord palsy occurs due of dysfunction of the recurrent laryngeal or vagus nerve, which leads to dysphonia. Patients with vocal cord palsy usually present with a fairly sudden onset of breathiness and weak, low-pitched dysphonia [2].

In clinical settings, various assessments have been performed to determine the severity of voice disorders. Voice evaluation is a widely used method that can easily quantify the quality of a patient’s voice disorders [11,12].

Auditory perceptual evaluation using the Grades of the Dysphonia Rating Scale (GRBAS), grade, roughness, breathiness, asthenia, strain, score, and consensus auditory perceptual evaluation of voice (cape-V) have been reliable measurements of each voice in many countries. In addition, acoustic and aerodynamic evaluations are relevant measurements in unilateral vocal cord palsy, which shows worse jitter, shimmer, noise/harmonic ratio, and maximum phonation time compared with normal voice [13,14].

Maryn et al. [3] described that CPP is one of the methods to measure voice quality, which has been qualified as the most promising and perhaps robust acoustic measure of breathy voice. The CPP value is a spectral representation of the voice spectrum, which is calculated by using the Fourier transformation of the voice spectrum. As CPP values are calculated by a double Fourier transformation and not by pitch tracking algorithms, as they do with noise-to-harmonic ratios, jitter, and shimmer values, they remain a robust measure for severely dysphonic voices [15-17].

As shown in this study, a normal voice without any diseases would have a well-defined harmonic structure with a strong cepstral peak compared to a breathy and hoarse voice, which is a poorly defined harmonic structure that leads to a weak cepstral peak. Similar to our study, Balasubramanium et al. designed a study on CPP in patients with unilateral vocal fold palsy and their results revealed lower CPP values in the clinical group than in the control group [2,18].

The results of this study showed that both indices have a high level of validity in discriminating between different GRBAS [19,20].

The CPP values in Praat were consistently significantly lower than those of the speech tool across all the data pools included in the analysis, which shows that there is a correlation between the CPP values achieved by the Praat and speech tools. The PV test reflects a more serious voice problem, which could contribute to determining the type of treatment.

Although the Multidimensional Voice Program is still the tool of choice worldwide by most researchers in Vocology, with much research done with Praat, its use will soon become more valuable [21].

With more studies on Praat being performed in the field of vocology and phonetics, this software program will be a valuable tool for clinicians to use as an indicator of voice quality.

The use of Praat has several advantages to using Praat. First, the program is free for anyone in need. Second, because it is software that can be downloaded from the internet, there is no need for additional space or equipment other than computers. Third, the pitch analysis algorithm is the most accurate in the world, as mentioned by Boersma and Weenink [10] which is the only linguistically oriented learning algorithm that can handle dynamic length changes (ejectives), non-glottal myo-eleastics (trills), and sucking effects (clicks and implosives); the gradual learning algorithm is the only linguistically oriented learning algorithm that can handle free variation. Forth, there is a possibility to use to decide the treatment modality to vocal cord palsy patients.

This study has several limitations. The first limitation is the small sample size. In only 30 patients with vocal cord palsy, CPP values did not represent all patients with vocal cord palsy. Second, the results of the study show a correlation between the two software programs, yet there is no certain numerical index that indicates whether a patient has a normal voice or a voice disorder. Therefore, further studies are needed to support and provide indices to validate people’s voice problems.

In conclusion, praat is an open-access software freely available and sufficiently used for voice evaluation in patients with vocal cord palsy. The data of this study supports that the values obtained by Praat and ST are positively correlated which can lead to further studies reflecting the severity of vocal cord palsy and could contribute to decide the type of treatment.

Supplementary Materials

Article information Continued