Cepstral Peak Sensitivity: A Theoretic Analysis and Comparison of Several Implementations

Abstract

Summary Objective The aim of this study was to develop a theoretic analysis of the cepstral peak, to compare several cepstral peak software programs, and to propose methods for reducing variability in cepstral peak estimation. Study Design Descriptive, experimental study. Methods The theoretic cepstral peak value of a pulse train was derived and compared to estimates computed for pulse train WAV files using available cepstral peak software programs: 1) Hillenbrand’s cepstral peak prominence (CPP) software, 2) KayPENTAX Multi-Speech implementation of CPP, and 3) a MATLAB implementation using cepstral interpolation. Cepstral peak variation was also investigated for synthetic breathy vowels. Results For pulse trains with period T samples, the theoretic cepstral peak is 1/2+ε/T, |ε|<0.1 for all pulse trains (ε=0 for integer T). For fundamental frequencies between 70 and 230 Hz, cepstral peak mean ± st. dev. was 0.496±0.002 using cepstral interpolation and 0.29±0.03 using Hillenbrand’s software, whereas CPP was 35.0±3.8 dB using Hillenbrand’s software and 20.5±2.7 dB using KayPENTAX’s software. CP and CPP vs. signal-to-noise ratio for synthetic breathy vowels were fit to a logistic model for the Hillenbrand (R2 = 0.92) and KayPENTAX (R2 = 0.82) estimators as well as an ideal estimator (R2 = 0.98) which used a period-synchronous analysis. Conclusions The findings indicate that several variables unrelated to the signal itself impact cepstral peak values, with some factors introducing large variability in cepstral peak values that would otherwise be attributed to the signal (e.g., voice quality). Variability may be reduced by using a period-synchronous analysis with Hann windows.

Impact of Instructed Laryngeal Manipulation on Acoustic Measures of Voice–Preliminary Results

“…A lower intensity may also result depending on vocal tract configuration which would also result in a lower CPP value, as found in our results. Although some authors have maintained that the effects of the vocal tract can be isolated from those of the voice source, 81 our findings do not support those claims. This characteristic of cepstral measurements must be taken into consideration if this analysis is to be used to assess vocal function.…”

Respiration Rate Estimation Based on Independent Component Analysis of Accelerometer Data: Pilot Single-Arm Intervention Study

“…The cepstrum was calculated through inverse Fourier transform. In this way, the harmonic component of signals could be acquired [ 19 , 20 ].…”

The relationship between biomechanics of pharyngoesophageal segment and tracheoesophageal phonation

“…CPP (in decibel, dB) 57 was measured using the Analysis of Dysphonia in Speech and Voice (ADSV) 58 and relative intensity (dB) was measured using the intensity analysis function. CPP was used as it has been recommended in the assessment of voice quality not suitable for more traditional analysis such as noise-to-harmonic ratio, jitter and shimmer 59 .…”