Cepstral Peak Prominence Research Articles

Kinematic modeling and acoustic measures of breathy voice.

A variety of disordered vocal fold structural and vibratory parameters have been identified in patients with breathy dysphonia, although the contribution of each parameter to voice quality is unknown. This study is part of an effort to determine the relation between production and voice quality. A kinematic vocal fold model [based on I. R. Titze, Speech Commun., 8, 191–201 (1989)] that emulates the medial surface of the vocal folds was used to ascertain factors that lead to breathy voice. Based on the ratings of four expert listeners, five parameters were selected for further study. Combinations of multiple levels of adduction, bulging, nodal point, epilaryngeal area, and phase were used to generate 180 samples of sustained /ɑ/ and /ɪ/. Seven acoustic analyses were performed for each sample: maximum flow declination rate, maximum area declination rate, two measures of H1-H2, two measures of cepstral peak prominence, and noise to harmonics ratio. All parameters led to systematic changes in acoustic measures reflecting increasing breathiness, with several inconsistencies highlighting the uniqueness of each measure. The H1-H2 measures were inconsistent with the other data. Results will be described in light of how each measure can inform researchers about production. [Work supported in part by NIH R01-DC04789.]

Evaluation of a simplified phonetic annotation scheme for disordered speech

Attempts to train a computer to mimic a vocal pathology expert's perception of perceivable voice problems have had limited success. A recent study successfully used a Cepstrum-based calculation (CPPs) to detect dysphonic speech, but it made significant false negative and false positive errors [1]. Appropriate training data could improve calculation accuracy, but logistical and legal issues from the medical domain render large amounts of detailed, consistent training data difficult to produce. A simple scheme has been developed based on labeling one point in time for each of a few vowels in a read statement. Calculations from [1] and [2], centered on neighborhoods of these vowels will augment the CPPs classifier. The annotation scheme will be evaluated with respect to its ability to assist a classifier to automatically predict perceptual labels [1] from data, and practicalities such as inter/intra-annotator agreement. 1. Heman-Ackah, Y.D., Heuer, R.J., Michael, D.D., Ostrowski, R., Horman, M., Baroody, M., Hillenbrand, J.M., and Sataloff, R.T. (2003). "Cepstral peak prominence: A more reliable measure of dysphonia," Annals of Otology, Rhinology, and Laryngology, 112, 324-333. 2. "Perceived Level of Noise by Mark VII and Decibels (E)" S. S. Stevens, J. Acoustical Soc. Am. v. 51(2, part 2) 1972. pp. 575-602.

The effects of linguistic experience on the perception of pathologically-disordered phonation

Ladefoged [Vocal Fold Physiology: Contemporary Research and Clinical Issues, College Hill Press, San Diego, 1983] wrote, what is a pathological voice quality in one language may be phonologically contrastive in another If one persons pathology is another persons phonemic phonation, then speakers of languages with phonation contrasts should treat pathologically-disordered and linguistically-relevant phonations in the same way. To test this, Gujarati (breathy versus modal contrast), Spanish (no breathiness) and English (allophonic breathiness) listeners participated in similarity-rating tasks. One used Mazatec (breathy, modal, and creaky phonation) stimuli, the other pathologically-disordered stimuli. Gujaratis treated pathologically-disordered stimuli and Mazatec stimuli in the same way. With both types of stimuli, Gujaratis based their judgments on H1−H2 (amplitude of the first harmonic minus the second) and mapped the stimuli into clusters with similar H1−H2 values across the two experiments. In both experiments, English listeners relied on cepstral peak prominence and H1−H2, and Spanish listeners relied on H1−H2 and H1−A1 (amplitude of first formant peak). English and Spanish listeners grouped both types of stimuli into clusters with similar values for the acoustic measures. However, English and Spanish listeners differed in how they weighed the importance of the different dimensions across the experiments. To English and Spanish listeners, pathologically-disordered phonations are subtly different from linguistically-relevant ones.

Periodicity estimation in synthesized phonation signals using cepstral rahmonic peaks

Aperiodicity in sustained phonation can result from temporal, amplitude and waveshape perturbations, turbulent noise, nonlinear phenomena and non-stationarity of the vocal tract. General measures of the periodicity of the voice signal are of interest in, for example, quantifying voice quality and in the assessment of pathological voice. High and low quefrency cepstral techniques are employed to supply an index of the degree of voice signal periodicity. In the high quefrency region, the first rahmonic is used to provide an indication of the periodicity of the signal. A new measure, SRA (sum of rahmonic amplitudes) – utilising all rahmonics in the cepstrum, is tested against synthesis data (six levels of random jitter, cyclic jitter, shimmer and random noise). In addition, an existing popular technique using the first rahmonic (cepstral peak prominence, CPP) is assessed with synthesis data for the first time. Both measures decrease with increasing aperiodicity levels of the glottal source, decreasing more noticeably for noise and random jitter than for shimmer and cyclic jitter. CPP is shown to be relatively f 0-independent; however, the index appears to be less sensitive when compared against SRA.

The Effect of Perceptual Training on Inexperienced Listeners' Judgments of Dysphonic Voice

The purpose of this study was (1) to determine whether changes in intra- and interrater reliability occur for inexperienced listeners' judgments of overall severity, roughness, and breathiness in dysphonic and normal speakers after 2 hours of listener training; and (2) to determine the acoustic bases of inexperienced listeners' judgments before and after training. Prospective, single group, pre- and postdesign. Thirty adult dysphonic and six normal speaker samples were selected from a database. Samples included 21 test stimuli and 15 training stimuli of both sustained vowels and connected speech. Sixteen inexperienced listeners judged all samples for overall severity, roughness, and breathiness using visual analog scales. Each listener provided pretraining ratings at baseline. Listeners were then trained using 15 anchor voice samples and 15 training stimuli. During training, listeners were provided with definitions of rating dimensions, accuracy feedback, and anchor samples. Listeners then judged test stimuli in a posttraining session. Speaker samples also were analyzed acoustically. Intrarater reliability was least variable for judgments of overall severity, but improved further with training. Listener judgments of roughness and breathiness in vowels were least reliable at baseline, but they significantly improved between listeners after training. Finally, measures of cepstral peak prominence significantly predicted all voice quality judgments except roughness in vowels, which was predicted by shimmer. The acoustic bases of group perceptual judgments did not seem to change with training. These findings have implications for developing training programs in perceptual evaluation and mapping relationships between acoustic and perceptual characteristics of voice disorders.

Reliability of calculating the cepstral peak without linear regression analysis

Summary Measures of cepstral peak prominence, using the smoothing algorithm and linear regression analysis software developed by Hillenbrand, have been shown to be reliable predictors of dysphonia in voice samples. 1–4 Recently, the Computerized Speech Laboratory [(CSL) Kay Elemetrics, Pinebrook, New Jersey] has introduced cepstral analysis as a component of that software package. The cepstral peak, in this instance, is calculated by the voice clinician analyzing the phonatory sample by subtracting the value of the peak from the apparent baseline signal. This study compares the ability of cepstral peak values calculated from the CSL software to predict dysphonia reliably with that of the values produced by the smoothing algorithm and linear regression analysis of Hillenbrand. The results of this study show that linear regression analysis is an important step in calculating the cepstral peak prominence, thus limiting the usefulness of software programs that do not employ this step.

Acoustic and Perceptual Parameters Relating to Connected Speech Are More Reliable Measures of Hoarseness than Parameters Relating to Sustained Vowels

This report investigates the correlations between acoustic parameters and the perception of hoarseness by trained listeners. Both sustained vowels and connected speech were examined. Fourteen acoustic parameters from samples of sustained vowels and 2 from connected speech were measured. The results show that jitter, shimmer and cepstral peak prominence (CPP) are correlated with the perception of hoarseness in sustained vowels. CPP is strongly correlated with the perception of hoarseness in connected speech. Some evidence is seen that perception of hoarseness in connected speech is more valid than the perception of hoarseness in sustained vowels. It is concluded that CPP for connected speech is a more valid objective measure of hoarseness than jitter, shimmer or CPP for sustained vowels and that perception of hoarseness may be most accurate in connected speech, rather than isolated vowels.

Cepstral peak prominence: a more reliable measure of dysphonia.

Quantification of perceptual voice characteristics allows the assessment of voice changes. Acoustic measures of jitter, shimmer, and noise-to-harmonic ratio (NHR) are often unreliable. Measures of cepstral peak prominence (CPP) may be more reliable predictors of dysphonia. Trained listeners analyzed voice samples from 281 patients. The NHR, amplitude perturbation quotient, smoothed pitch perturbation quotient, percent jitter, and CPP were obtained from sustained vowel phonation, and the CPP was obtained from running speech. For the first time, normal and abnormal values of CPP were defined, and they were compared with other acoustic measures used to predict dysphonia. The CPP for running speech is a good predictor and a more reliable measure of dysphonia than are acoustic measures of jitter, shimmer, and NHR.

Objective acoustic and aerodynamic measures of breathiness in paralytic dysphonia.

This study was designed to determine which objective acoustic or aerodynamic parameters allowed a homogeneous group of patients with unilateral vocal fold paralysis (UVFP) to be distinguished from an age-matched and smoking-matched control group and to search for linear correlations between the objective parameters and the subjective breathiness ratings. Eight patients with recent-onset UVFP and 12 controls were prospectively studied. The acoustic parameters measured for the vowel /a/ at a comfortable frequency and intensity were: jitter, shimmer, harmonics-to-noise ratio, cepstral peak prominence, the difference between the levels of the first two harmonics and the relative energy above 6 kHz. Aerodynamic parameters included the mean flow rate during a sustained /a/ and intraoral pressure during the production of the phoneme /pi/. The long-term average spectrum was calculated for 40 s of text, and the relative average energies in four frequency bands were compared. Six judges rated a mid-/a/ sample using a five-parameter scale with four levels of severity. Nonparametric statistical analysis revealed significant differences ( P<.05) between the UVFP group and the control group for 14 of the 19 parameters studied. Correlations between the objective parameters and perceived breathiness differed in the two groups. Correlations were not always as expected as based on previous literature reports. These measurements provided an objective qualification of voice in patients with UVFP and successfully distinguished them from the normal controls. The objective acoustic and aerodynamic measurements had generally low linear correlations with breathiness ratings in the control group. Higher correlations were seen in the UVFP group, in which breathiness was best correlated with airflow measurements.

The timing of nonmodal phonation in vowels

In a language where breathiness or laryngealization is a contrastive property of vowels, such nonmodal phonation lasts longer and is more differentiated from modal phonation than in a language where nonmodal phonation results from the influence of preceding consonants. In Tagalog, breathy phonation occurs incidentally on vowels after /h/, and laryngealized phonation occurs after glottal stops. Mazatec, on the other hand, employs breathy and laryngealized vowels as separate phonemes that contrast with modal vowels. Several acoustic measures show that nonmodal and modal vowels are differentiated more strongly and over a longer duration in Mazatec than in Tagalog.An experiment examined words from six male and six female speakers of each of those languages, with corroborating modal and breathy vowels from four male and four female speakers of Chong, and modal and laryngealized vowels from one male speaker of Mpi. From each speaker, three vowels of each phonation type were analyzed. To determine the time course of phonation effects, measurements were made at 25 ms intervals through each vowel. The measurements were the amplitude differences between the first and second harmonic and between the first harmonic and the second formant, and cepstral peak prominence (a measure of periodicity).

The Relationship Between Cepstral Peak Prominence and Selected Parameters of Dysphonia

Traditional measures of dysphonia vary in their reliability and in their correlations with perceptions of grade. Measurements of cepstral peak prominence (CPP) have been shown to correlate well with perceptions of breathiness. Because it is a measure of periodicity, CPP should also predict roughness. The ability of CPP and other acoustic measures to predict overall dysphonia and the subcategories of breathiness and roughness in pathological voice samples is explored. Preoperative and postoperative speech samples from 19 patients with unilateral recurrent laryngeal nerve paralysis who underwent operative intervention were analyzed by trained listeners and by measures of smoothed CPP (CPPS), noise-to-harmonic ratio (NHR), amplitude perturbation quotient (APQ), relative average perturbation (RAP), and smoothed pitch perturbation quotient (sPPQ). The data were analyzed with bivariate Pearson correlation statistics. Grade of dysphonia and breathiness ratings correlated better with measurements of CPPS than with the other measures. CPPS from samples of connected speech (CPPS-s) best predicted overall dysphonia. None of the measures were useful in predicting roughness.

Objective Voice Quality Analysis Before and After Onset of Unilateral Vocal Fold Paralysis

This study was designed to investigate objective voice quality measurements in unilateral vocal fold paralysis (UVFP) by eliminating intersubject variability. To our knowledge this is the first report objectively analyzing paralytic dysphonia as compared to the same voice before onset of UVFP. The voices of two male subjects were prospectively recorded before and after the onset of iatrogenic UVFP (thoracic surgery). The following acoustic measurements of the vowel /A/ were performed using the CSL and MDVP (Kay Elemetrics): jitter, shimmer, harmonics-to-noise ratio, cepstral peak prominence, the relative energy levels of the first harmonic, the first formant and the third formant, the spectral slope in the low-frequency zone (0–1 kHz and 0–2 kHz), and the relative level of energy above 6 kHz. Distribution of spectral energy was analyzed from a long-term average spectrum of 40 seconds of text. Laryngeal aerodynamic measurements were obtained for one patient before and after onset of paralysis using the Aerophone II (Kay Elemetrics). Pitch and amplitude perturbation increased secondary to UVFP, while the harmonics-to-noise ratio and the cepstral peak prominence decreased. A relative increase in the mid-frequency and high-frequency ranges and a decrease in the low-frequency spectral slope were observed. Mean airflow rate and intraoral pressure increased, and glottal resistance and vocal efficiency decreased secondary to UVFP. The findings of this self-paired study confirm some but not all the results of previous studies. Measures involving the fundamental and the formants did not corroborate previous findings. Further investigation with vocal tract modeling is warranted.

Acoustic and perceptual evaluation of laryngeal reinnervation by ansa cervicalis transfer.

To evaluate the acoustic and perceptual results of laryngeal reinnervation with ansa cervicalis to recurrent laryngeal nerve anastomosis. Retrospective study of voice samples from 12 patients with unilateral recurrent laryngeal nerve paralysis, treated with ansa cervicalis to recurrent laryngeal nerve anastomosis. Samples were recorded before surgery and at least 8 months after surgery. The samples were subjected to several acoustic analyses sensitive to paralytic dysphonia, including cepstral peak prominence, noise-to-harmonics ratio, and measures of frequency and amplitude perturbation. The voice samples from the patients were randomized with age- and sex-matched samples from normal subjects and judged by trained listeners for overall dysphonia, roughness, breathiness, asthenia, and strain. The preoperative and postoperative results were compared statistically, and the postoperative results were compared with the matched normal subjects. As a group, the patients showed improvement (P < .05) in cepstral peak prominence, frequency perturbation, and perceptual judgments of overall dysphonia, breathiness, and asthenia. The best results occurred in patients with isolated vocal fold paralysis. The postoperative group as a whole did not improve to the level seen in matched normals. Suboptimal results were seen primarily in patients with untreated laryngeal or extralaryngeal pathology beyond the laryngeal paralysis. These data suggest that laryngeal reinnervation has the potential to bring about a return to normal or near-normal voice in patients with isolated unilateral vocal fold paralysis.

Acoustic correlates of dysphonia: type and severity

The purpose of this study was to explore the acoustic discrimination and graded severity of three clinical voice types. Listeners classified 102 samples of dysphonic vowels /a/ and /i/ on the basis of voice types: breathy, hoarse, and strained. The vowels were analyzed acoustically with two measures of perturbation and 2 measures of spectral noise. Discriminant analysis showed that apriori, acoustic classifications of voice type were made with 92% accuracy using four acoustic parameters: (a) cepstral peak prominence (CPP), (b) jitter standard deviation (SD-J), (c) fundamental frequency (F 0), and (d) standard deviation of signal-to-noise ratio (SD-SNR). Findings suggest that voice type is associated with the interaction of spectral noise, fundamental frequency, and signal irregularity, and that dysphonic severity is associated with similar parameters, regardless of voice type.