Cepstral Measures Research Articles

Cepstral Characteristics of Voice in Indian Female Classical Carnatic Singers

The cepstral measures are gaining attention in the voice analysis as it is reported to be strong predictor of voice impairments. As singer's voice is reported to be present with more harmonics, it can be hypothesized that there would be an increase in the value of cepstral parameters. Also, little is known about the cepstral characteristics of voices in Indian classical Carnatic singers; and hence, the present study was undertaken with an aim to characterize the singer's voice using cepstral measures. Thirty Indian female classical Carnatic singers and 30 nonsingers participated in the study. Phonation of vowel /a/ was recorded at their habitual pitch and loudness. The cepstral peak prominence and smoothened cepstral peak prominence were analyzed using the Hillenbrand algorithm available in Speech Tool software. Mean and standard deviation was calculated for each group. Obtained scores were analyzed using independent t test. Mean raw scores were found to be increased in classical Carnatic singers in comparison with nonsingers. Results of independent t test revealed significant difference between the means of singers and nonsingers for both the cepstral measures at P<0.05. The present study investigated the cepstral characteristic of voice in singers and nonsingers. Results revealed that there was a significant difference between the means of singers and nonsingers indicating that cepstral parameters were higher among the singers in comparison with nonsingers. The obtained results were attributed to the harmonic organization in the voices of singers.

Cepstral analysis of voice in healthy aged individuals

Introduction: The anatomical and physiological changes in the phonatory system due to aging have a major impact on voice. Widely researched parameters of voice such as fundamental frequency and the perturbation measures depend on the location of the exact pitch pulses, and it may yield unreliable results in case of a severely aperiodic voice. Aim of the Study: The present study focused on determining the cepstral characteristics of voice in healthy aging individuals. Methods: All the participants among the young adults, middle-aged adults, and old adults were required to sustain the vowel/a/. Voice samples were analyzed using Hillenbrand's Z tool software to obtain the cepstral measures. Results: Results showed that there was an increase in the cepstral measures with aging, thus suggesting that the harmonic structure of voice is not affected in healthy aged individuals. Gender also influences the cepstral measures in healthy aging population. Conclusion: The results of the present study could be used by the voice clinicians while assessing adults with voice disorders. It will also help the clinicians in delineating the age and gender differences in the vocal parameters.

Exploring the Relationship Between Spectral and Cepstral Measures of Voice and the Voice Handicap Index (VHI)

The purpose of this study was to examine the strength of relationship between impairment-level acoustic measures derived from spectral- and cepstral-based analyses (including the cepstral peak prominence [CPP]; ratios of low vs high frequency spectral energy; and the respective standard deviations [SDs] for these measures) and a disablement measure (the total Voice Handicap Index [VHI] score) in a large and diverse group of voice-disordered and control subjects. The relationship between total VHI and the Cepstral Spectral Index of Dysphonia (CSID-a multivariate estimate of dysphonia severity) was also examined. Subjects were 332 adults (116 males and 216 females) comprised of voice-disordered subjects who presented to a physician with a voice-related complaint (n=258) and a group of nonvoice-disordered control subjects (n=74). A VHI 30-item score and speech/voice samples including the second and third sentences of The Rainbow Passage and productions of the sustained vowel /ɑ/ were obtained for each subject. Sentence and sustained vowel samples were analyzed using the Analysis of Dysphonia in Speech and Voice (ADSV) program (ADSV model 5109 v.3.4.2; KayPENTAX, Montvale, NJ). Across all subjects, low-to-moderate strength Spearman rho (rs) correlations were observed between the total VHI and the CPP and the CSID in both speech and vowel contexts and for the CPP SD from continuous speech (rs's ranging from -0.45 to -0.49 for VHI vs CPP; 0.47 for VHI vs CSID; -0.44 for VHI vs CPP SD). Several other measures obtained from spectral or cepstral analyses also were observed to correlate with total VHI, although increased variability in the strength, direction, and overall significance of these other variables was observed depending on gender and elicited context. Voice-related disablement occurs within a context. In contrast, impairment-level measures of phonatory function (like the spectral and cepstral measures included in this study) are by nature decontextualized and appear to correlate low-to-moderately with quality of life measures like the VHI. Therefore, spectral and cepstral acoustic measures and the VHI should be viewed as providing relatively unique, meaningful, and complementary information.

Cepstral Characteristics of Voice before, during and after Menopause: A Cross-sectional Study

ABSTRACT Background Previous studies on effects of menopause on voice characteristics have not compared the voice characteristics before, during and after menopause. It is also possible that these voice changes cannot be captured by traditional acoustic measures such as frequency and amplitude perturbation measures. Hence a robust acoustic tool such as cepstrum was chosen to capture these vocal changes before, during and after menopause and hence the present study was attempted. Method The present study followed a cross-sectional study design. The participants consisted of 25 premenopausal women, 25 perimenopausal women and 25 postmenopausal women. Sustained production of/a/ and narration was recorded using Z tool and cepstral peak prominence (CPP) and smoothened cepstral peak prominence (CPPS) was analyzed. Results The results of one-way ANOVA revealed that there was a significant main effect of the group at p &lt; 0.05 for all the cepstral measures targeted in the study. Post hoc analysis indicated that there was no significant difference between pre and perimenopausal women for all the cepstral measures at p &lt; 0.05. However, there was a significant difference between pre and postmenopausal women and peri and postmenopausal women at p &lt; 0.05. These results suggests that postmenopausal women had overall good voice quality than pre and perimenopausal women. Conclusion Results indicated that post menopausal women had increased cepstral measures when compared to pre and perimenopausal women. Further, studies are required to correlate laryngeal examination and hormonal levels with that of the voice changes before, during and after menopausal women. How to cite this article Balasubramanium RK, Bhat JS. Cepstral Characteristics of Voice before, during and after Menopause: A Cross-sectional Study. Int J Phonosurg Laryngol 2014;4(2):50-54.

Acoustic discrimination of Parkinsonian speech using cepstral measures of articulation

The effects of idiopathic Parkinson’s disease (IPD) on speech include articulatory imprecision. We quantified articulation rate and range acoustically using cepstral coefficients to represent vocal tract settings. Cepstral coefficients were extracted from 10 sentences spoken by 76 talkers, half of which were diagnosed with IPD. Articulation range was estimated for each sentence as the sum across cepstral coefficients of the standard deviation of each coefficient, and articulation rate was estimated using the same procedure, replacing cepstral coefficients with delta coefficients. The mean ± standard deviation (N = 380 sentences) for the articulation measures of range (7.95 ± 0.50 vs. 6.66 ± 0.53) and rate (5.64 ± 0.56 vs. 4.40 ± 0.46) were significantly different (t-test, p &amp;lt; 0.001) for normal vs. IPD speech, respectively. In a leave-one-talker-out classification experiment, range accuracy was 90.1%, rate accuracy was 88.8%, and accuracy was 92.9% using a combined model of articulation range and rate. The strengths of the articulation measures include 1) sensitivity to IPD speech, 2) reliance on cepstral coefficients which have been used for over 30 years to represent speech, 3) no segmentation requirements, 4) low sensitivity to speech material, and 5) effective with only 2 seconds of speech.

Spectral- and Cepstral-Based Measures During Continuous Speech: Capacity to Distinguish Dysphonia and Consistency Within a Speaker

Spectral- and cepstral-based acoustic measures are preferable to time-based measures for accurately representing dysphonic voices during continuous speech. Although these measures show promising relationships to perceptual voice quality ratings, less is known regarding their ability to differentiate normal from dysphonic voice during continuous speech and the consistency of these measures across multiple utterances by the same speaker. The purpose of this study was to determine whether spectral moments of the long-term average spectrum (LTAS) (spectral mean, standard deviation, skewness, and kurtosis) and cepstral peak prominence measures were significantly different for speakers with and without voice disorders when assessed during continuous speech. The consistency of these measures within a speaker across utterances was also addressed. Continuous speech samples from 27 subjects without voice disorders and 27 subjects with mixed voice disorders were acoustically analyzed. In addition, voice samples were perceptually rated for overall severity. Acoustic analyses were performed on three continuous speech stimuli from a reading passage: two full sentences and one constituent phrase. Significant between-group differences were found for both cepstral measures and three LTAS measures (P<0.001): spectral mean, skewness, and kurtosis. These five measures also showed moderate to strong correlations to overall voice severity. Furthermore, high degrees of within-speaker consistency (correlation coefficients ≥0.89) across utterances with varying length and phonemic content were evidenced for both subject groups.

Spectral, cepstral, and multivariate exploration of tracheoesophageal voice quality in continuous speech and sustained vowels

The quality of tracheoesophageal voice can vary substantially. Although previous research has identified acoustic differences between various types of voicing (i.e., laryngeal, tracheoesophageal, esophageal, etc.), acoustic analysis has failed to quantify the degree of alaryngeal voice quality. This study assessed the value of several cepstral, spectral, and perturbation measures in quantifying the overall quality of tracheoesophageal voice production. Cross-sectional, correlational. Continuous speech and sustained vowel samples from 16 tracheoesophageal speakers were concatenated and perceptually rated in a paired comparison paradigm on overall voice quality by four experienced clinicians. After removing the nonvoiced fragments within the continuous speech samples, the concatenated samples were analyzed with 47 perturbation, spectral, and cepstral measures. Correlation between perceptual ratings and acoustic measures was assessed. Multiple regression analysis resulted in a two-factor acoustic model for the measurement of overall voice quality of the concatenated samples. The reliability of the perceptual judgements was moderate to high. The prominence of the cepstral peak (CPP) and of the first two spectral harmonics appeared to be the strongest correlates of tracheoesophageal voice quality. A linear regression-based combination of CPP and the height of the second harmonic produced a correlation of 0.87 with listener judgments. It is clinically feasible to investigate both continuous speech and sustained vowel samples of tracheoesophageal speakers with acoustic methods described and assessed in this report. Results are discussed in the context of existing literature.

Objective quality evaluation in blind source separation for speech recognition in a real room

The determination of quality of the signals obtained by blind source separation is a very important subject for development and evaluation of such algorithms. When this approach is used as a pre-processing stage for automatic speech recognition, the quality measure of separation applied for assessment should be related to the recognition rates of the system. Many measures have been used for quality evaluation, but in general these have been applied without prior research of their capabilities as quality measures in the context of blind source separation, and often they require experimentation in unrealistic conditions. Moreover, these measures just try to evaluate the amount of separation, and this value could not be directly related to recognition rates. Presented in this work is a study of several objective quality measures evaluated as predictors of recognition rate of a continuous speech recognizer. Correlation between quality measures and recognition rates is analyzed for a separation algorithm applied to signals recorded in a real room with different reverberation times and different kinds and levels of noise. A very good correlation between weighted spectral slope measure and the recognition rate has been verified from the results of this analysis. Furthermore, a good performance of total relative distortion and cepstral measures for rooms with relatively long reverberation time has been observed.

On the use of bandpass liftering in speech recognition

In a template-based speech recognition system, distortion measures that compute the distance or dissimilarity between two spectral representations have a strong influence on the performance of the recognizer. Accordingly, extensive comparative studies have been conducted to determine good distortion measures for improved recognition accuracy. Previous studies have shown that the log likelihood ratio measure, the likelihood ratio measure, and the truncated cepstral measures all gave good recognition performance (comparable accuracy) for isolated word recognition tasks. In this paper we extend the interpretation of distortion measures, based upon the observation that measurements of speech spectral envelopes (as normally obtained from standard analysis procedures such as LPC or filter banks) are prone to statistical variations due to window position fluctuations, excitation interference, measurement noise, etc., and may not accurately characterize the true speech spectrum because of analysis model constraints. We have found that these undesirable spectral measurement variations can be partially controlled (i.e., reduced in the level of variation) by appropriate signal processing techniques. In particular, we have found that a bandpass process reduces the variability of the statistical components of LPC-based spectral measurements and hence it is desirable to use such a liftering process in a speech recognizer. We have applied this liftering process to several speech recognition tasks: in particular, single frame vowel recognition and isolated word recognition. Using the liftering process, we have been able to achieve an average digit error rate of 1 percent in a speaker-independent isolated digit test. This error rate is about one-half that obtained without the liftering process.