Analysis Of Pathological Voices Research Articles

Quest for Speech Enhancement Method in the Analysis of Pathological Voices

Quest for Speech Enhancement Method in the Analysis of Pathological Voices

Acoustic Voice Analysis of Normal and Pathological Voices in Indian Population Using Praat Software.

Acoustic voice analysis is still a valuable technique which enables voice clinicians to compare voices to differentiate them into normal and abnormal. The present study was undertaken to standardize acoustic voice parameters in normal healthy adult individuals and gender comparison among them and also acoustic voice analysis of pathological voices and it's comparison with normal healthy voices. Voice samples of vowels /a/, /i/ and /u/ of 80 normal healthy adults (males = 40, females = 40) of control group and 40 patients with dysphonic voice of case group collected and acoustic voice parameters were extracted by using Praat software. There were statistically significant higher values of fundamental frequency (F0) in females, while jitter local (%), shimmer local (%) and harmonic to noise ratio (HNR) had no gender differences in normal healthy voices. Pathological voices of case group subjects with laryngeal pathologies had statistically significant higher values of jitter local (%), shimmer local (%) and lower values of HNR as compare to normal healthy voices of control group. Objective voice analysis by using Praat software is convenient, reliable and cost effective method. This study establishes normative acoustic voice parameters in normal healthy adults. There are no gender differences in adult healthy voices except fundamental frequency (F0), which is higher in females. Patients who are with dysphonic voices due to laryngeal pathologies had altered values of acoustic parameters compared to normophonic adults and clinicians can precisely differentiate pathological voices from normophonics.

Effect of vowel context in cepstral and entropy analysis of pathological voices

This study investigates the effect of vowel context (excerpted from speech versus sustained) on two voice quality measures: the cepstral peak prominence smoothed (CPPS) and sample entropy (SampEn). ...

An incremental method combining density clustering and support vector machines for voice pathology detection

Machine learning techniques are a valuable tool for discriminative classification. They have been applied to a diverse range of applications in speech processing, such as the analysis of pathological voices. We propose, in this paper, a novel policy, called Incremental DBSCAN-SVM in order to detect noises, to analyze and to classify pathological voice from normal voice. We use a modified density-based clustering algorithm named DBSCAN with an incremental learning in order to detect noisy samples. Then, the output model is submitted to Support Vector Machines (SVM) classifier with a Radial Basis Function (RBF) kernel to discriminate between normal and pathological voices. Our method has the ability to handle incremental and dynamic voices database which evolve over time. We support our approach with empirical evaluation using voices data set from the Massachusetts Eye and Ear Infirmary Voice and Speech Laboratory (MEEI) database to show the effectiveness of our method in terms of detection voice disorders.

Quantitative analysis of pathological voice and identification with artificial neural network

Objective:To discuss the characteristic of the clinical pathological voice and the feasibility of computer automatic identification of pathological voice.Method:A total of 129 clinical patients with polyp of vocal cord were selected as the pathological voice group, while a total of 125 people with normal voice were selected from the community as the control group. Praat software was used to collect and analyze the related acoustic parameter values of two groups of cases, including Jitter, Shimmer, harmonic to noise ratio (HNR), signal to noise ratio (SNR), and normalized noise energy (NNE). Pathological voice group and control group were used as training set and testing set for neural network testing, and another 140 cases of pathological voice and normal voice data were selected as a validation set. SPSS Modeler was used for artificial neural network reconstruction to calculate the identification rate of pathological voice. Result:This study found according to the calculation of groups with different genders that Jitter, Shimmer and NNE were increased in pathological voice group compared with the normal group (P< 0.05), while HNR and SNR were decreased compared with the normal group (P< 0.05). Recognition rate of artificial neural network model on pathological voice is 75.7%.Conclusion:Objective voice analysis is helpful in the identification of pathological voice. Artificial neural network has higher accuracy in recognition of pathological voice, with good clinical application value.

Acoustic Voice Analysis Using the Praat programme: Comparative Study With the Dr. Speech Programme

Introduction and objectivesThe European Laryngological Society (ELS) basic protocol for functional assessment of voice pathology includes 5 different approaches: perception, videostroboscopy, acoustics, aerodynamics and subjective rating by the patient. In this study we focused on acoustic voice analysis.The purpose of the present study was to correlate the results obtained by the commercial software Dr. Speech and the free software Praat in 2 fields:1. Narrow-band spectrogram (the presence of noise according to Yanagihara, and the presence of subharmonics) (semi-quantitative).2. Voice acoustic parameters (jitter, shimmer, harmonics-to-noise ratio, and fundamental frequency) (quantitative). Materials and methodsWe studied a total of 99 voice samples from individuals with Reinke's oedema diagnosed using videostroboscopy. One independent observer used Dr. Speech 3.0 and the second one used the Praat programme (Phonetic Sciences, University of Amsterdam).The spectrographic analysis consisted of obtaining a narrow-band spectrogram from the previous digitalised voice samples by the 2 independent observers. They then determined the presence of noise in the spectrogram, using the Yanagihara grades, as well as the presence of subharmonics. As a final result, the acoustic parameters jitter, shimmer, harmonics-to-noise ratio and fundamental frequency were obtained from the 2 acoustic analysis programmes. ResultsThe results indicated that the sound spectrogram and the numerical values obtained for shimmer and jitter were similar for both computer programmes, even though type 1, 2 and 3 voice samples were analysed. ConclusionsThe Praat and Dr. Speech programmes provide similar results in the acoustic analysis of pathological voices.

Análisis acústico de la voz mediante el progama Praat: estudio comparativo con el programa Dr. Speech

Introduction and objectivesThe European Laryngological Society (ELS) basic protocol for functional assessment of voice pathology includes 5 different approaches: perception, videostroboscopy, acoustics, aerodynamics and subjective rating by the patient. In this study we focused on acoustic voice analysis.The purpose of the present study was to correlate the results obtained by the commercial software Dr. Speech and the free software Praat in 2 fields:1. Narrow-band spectrogram (the presence of noise according to Yanagihara, and the presence of subharmonics) (semi-quantitative).2. Voice acoustic parameters (jitter, shimmer, harmonics-to-noise ratio, fundamental frequency) (quantitative). Material and methodsWe studied a total of 99 voice samples from individuals with Reinke's oedema diagnosed using videostroboscopy. One independent observer used Dr. Speech 3.0 and a second one used the Praat program (Phonetic Sciences, University of Amsterdam).The spectrographic analysis consisted of obtaining a narrow-band spectrogram from the previous digitalised voice samples by the 2 independent observers. They then determined the presence of noise in the spectrogram, using the Yanagihara grades, as well as the presence of subharmonics. As a final result, the acoustic parameters of jitter, shimmer, harmonics-to-noise ratio and fundamental frequency were obtained from the 2 acoustic analysis programs. ResultsThe results indicated that the sound spectrogram and the numerical values obtained for shimmer and jitter were similar for both computer programs, even though types 1, 2 and 3 voice samples were analysed. ConclusionsThe Praat and Dr. Speech programs provide similar results in the acoustic analysis of pathological voices.

A hybrid expert system for automatic detection of voice disorders

Pathological voice analysis is a challenging task and an important area of research in voice disorder identification. Until now, the long-time acoustic (LTA) parameters are used primitively to classify the disordered voices into pathological and normal. Selection of such optimal LTA features is a disputing task. Previous researchers have used various data projection methods like principle component analysis (PCA), linear discriminant analysis (LDA) and sub-optimal searching techniques like sequential forward selection (SFS), sequential backward selection (SBS), and individual feature selection (IFS) methods for this purpose. But, these methods work efficiently for linearly separable datasets only. In order to overcome these issues, we propose a hybrid expert system in this paper, which includes the optimal selection of LTA parameters using genetic algorithm (GA), followed by non-linear classification algorithms to classify the two classes of voice samples. Nowadays, though many non-linear and high-order spectral parameters of voices have been used in this application, LTA features are scoring more importance because their clinical diagnosis is of more ease. Within this context, the GA-based feature vector quantisation combined with SVM classification is demonstrated to be more reliable, yielding 96.86% of classification accuracy for a feature vector of length 10.

Classification of Normal and Pathological Voice using GA and SVM

analysis of pathological voice is a challenging and an important area of research in speech processing. Acoustic characteristics of voice are used mainly to discriminate normal voices from pathological voices. This study explores methods to find the ability of acoustic parameters in discrimination of normal voices from pathological voices. An attempt is made to analyze and to classify pathological voice from normal voice in children. The classification of pathological voice from normal voice is implemented using support vector machine (SVM). The normal and pathological voices of children are used to train and test the classifier. A dataset is constructed by recording speech utterances of a set of Tamil phrases. The speech signal is then analyzed in order to extract the acoustic parameters such as the Signal Energy, pitch, formant frequencies, Mean Square Residual signal, Reflection coefficients, Jitter and Shimmer. In this study various acoustic features are combined to form a feature set, so as to detect voice disorders in children based on which further treatments can be prescribed by a pathologist. A Genetic Algorithm (GA) based feature selection is utilized to select best set of features which improves the classification accuracy.

Support vector wavelet adaptation for pathological voice assessment

The presence of abnormalities in the vocal system affects the quality of the voice and changes its characteristics. Digital analysis of pathological voices can be an effective and non-invasive tool for the detection of such alterations. This paper proposes a wavelet-based method to distinguish between normal and disordered voices. Wavelet filter banks are used in conjunction with support vector machines, as feature extractors and classifiers, respectively. Orthogonal filter banks are implemented using a highly efficient structure known as “lattice” that parameterizes filter banks and produces a few parameters. The overall problem is to find these parameters such that perfect classification is achieved. To search for such parameters, a genetic algorithm with a fitness function corresponding to the classification result is applied. Simulation is done on the KAY database (a comprehensive database including 710 normal and pathological voice signals, developed by the Massachusetts Eye and Ear Infirmary Voice and Speech Lab), and one additional test set. It is observed that a genetic algorithm is able to find the filter bank parameters such that a 100% correct classification rate is achieved in classifying normal and pathological voices when the test is performed on both databases.

An efficient approach using HOS-based parameters in the LPC residual domain to classify breathy and rough voices

Although a considerable number of studies have been focused on the analysis of pathological voices using conventional parameters such as jitter, shimmer, and signal-to-noise ratio (SNR), these parameters have been found to be sensitive to variations in pitch extraction algorithm and cannot analyze severely disordered voice signals which exhibit irregular or aperiodic waveforms. In this paper, higher-order statistics (HOSs) analysis, which is independent of pitch period, is derived from linear predictive coding (LPC) residuals to describe breathy and rough voices. Recordings of a sustained /a/ from 23 individuals with breathy voices and 30 individuals with rough voices were collected from the disordered voice database distributed by the Japanese Society of Logopedics and Phoniatrics. We extracted conventional parameters as well as the HOS-based parameters such as the normalized skewness and the normalized kurtosis. On the other hand, we calculated HOS-based parameters from the LPC residual domain. The results showed that the HOS-based parameters and the HOS-based parameters estimated from the LPC residual are different for rough and breathy voices. Conventional parameters were not distinctive for these voices. Classification and regression tree (CART) was used to combine multiple parameters and to classify breathy and rough voices. Using the HOS-based parameters, the CART achieved an accuracy of 85.0% with the optimal decision tree generated by means of the normalized skewness and kurtosis. When the HOS-based parameters using LPC residual were used, the optimal decision tree was 88.7% accurate and the variances of the normalized skewness and kurtosis were included.

Objective Acoustic Analysis of Pathological Voices from Patients with Vocal Nodules and Polyps

Objective: We aim to examine the abilities of objective acoustic analysis methods (nonlinear dynamic and traditional perturbation measures) to describe voices from individuals with vocal nodules and polyps. Subjects and Methods: Sustained vowel recordings from normal subjects, patients with vocal nodules, and patients with vocal polyps were analyzed. Perturbation measures of jitter and shimmer were obtained with the Multi-Dimensional Voice Program (MDVP) and CSpeech. Signal-to-noise ratio was calculated using CSpeech. Nonlinear dynamic measures of phase space reconstruction and correlation dimension were also applied to analyze the voices. Results: A significant difference between normal and polyp groups was found in jitter and shimmer obtained from MDVP, as well as in jitter and signal-to-noise ratio from CSpeech. However, no parameters significantly differentiated between normal and nodule groups. Shimmer from CSpeech did not reveal any significant differences among any of the groups. Correlation dimension values for the nodule and polyp groups were significantly higher than the normal group. Conclusion: Nonlinear dynamic analysis has great potential value for the characterization of voice from patients with vocal nodules and polyps. The combination of traditional perturbation and nonlinear dynamic measures may improve our ability to provide objective clinical analysis of voices with vocal mass lesions.

Multidimensional acoustic analysis of pathological voice

There are subjective and objective ways to examine the effects of vocal therapy in voice disorders. The most precise and objective check-up is the use of computer voice analysis. The aim of the research was to perform a detailed analysis of acoustic structure of the vowel A before and after voice treatment in patients with vocal fold nodules in order to obtain objective verification of the vocal rehabilitation success. We examined 30 female patients, aged 34.6 +/- 6.69 years, with vocal fold nodules. Acoustic parameters of voice were compared with the control group consisting of 21 subjects without voice pathology. In all persons the vowel A was recorded and analyzed before and after a month of vocal therapy. The success of the vocal therapy was tracked using computer analysis of vocal structure. Signal, noise and tremor parameters were processed. Of the analyzed vowel A parameters: STD, PER, JITA, JITT, RAP, vFO, ShdB, SHIM, APQ, VTI, SPI, FO, NHR, FTRI, eleven improved (p < 0.05 and p < 0.01). Three parameters (FO, NHR, FTRI) changed showing improvement, but the obtained differences were not statistically significant (p > 0.05). Based on the obtained results it was concluded that vocal therapy gave satisfactory results, but that it should be continually applied until full stabilization of the voice.

Pitch Deviation Analysis of Pathological Voice in Connected Speech

This study compares normal and pathologic voices using a novel voice analysis algorithm that examines pitch deviation during connected speech. The study evaluates the clinical potential of the algorithm as a mechanism to distinguish between normal and pathologic voices using connected speech. Adult vocalizations from normal subjects and patients with known benign free-edge vocal fold lesions were analyzed. Recordings had been previously obtained in quiet under controlled conditions. Two phrases and sustained /a/ were recorded per subject. The subject populations consisted of 10 normal and 31 abnormal subjects. The voice analysis algorithm generated 2-dimensional patterns that represent pitch deviation in time and under variable window widths. Measures were collected from these patterns for window widths between 10 and 250 ms. For comparison, jitter and shimmer measures were collected from sustained /a/ by means of the Computerized Speech Lab (CSL). A t-test and tests of sensitivity and specificity assessed discrimination between normal and abnormal populations. More than 58% of the measures collected from connected speech outperformed the CSL jitter and shimmer measures in population discrimination. Twenty-five percent of the experimental measures (including /a/) indicated significantly different populations (p < .01%). The results demonstrate that the algorithm distinguishes between normal and abnormal populations by use of samples of connected speech.

Wavelet-based denoising for improving nonlinear dynamic analysis of pathological voices

A pre-processing scheme based on wavelet denoising is applied to improve nonlinear dynamic analysis for pathological voices. Nonlinear dynamic analyses are corrupted by high noises, yet can be effectively improved using this novel voice analysis scheme. The estimated correlation dimensions of the noisy pathological voices are statistically higher than those of the original voices; however, there is no significant difference between the dimensions of the denoised voices and the original voices. Nonlinear dynamic analysis via wavelet denoising represents a valuable method for describing pathological human voices under noisy conditions to assess laryngeal pathologies.

The correlogram: a visual display of periodicity.

Fundamental frequency (F0) extraction is often used in voice quality analysis. In pathological voices with a high degree of instability in F0, it is common for F0 extraction algorithms to fail. In such cases, the faulty F0 values might spoil the possibilities for further data analysis. This paper presents the correlogram, a new method of displaying periodicity. The correlogram is based on the waveform-matching techniques often used in F0 extraction programs, but with no mechanism to select an actual F0 value. Instead, several candidates for F0 are shown as dark bands. The result is presented as a 3D plot with time on the x axis, correlation delay inverted to frequency on the y axis, and correlation on the z axis. The z axis is represented in a gray scale as in a spectrogram. Delays corresponding to integer multiples of the period time will receive high correlation, thus resulting in candidates at F0, F0/2, F0/3, etc. While the correlogram adds little to F0 analysis of normal voices, it is useful for analysis of pathological voices since it illustrates the full complexity of the periodicity in the voice signal. Also, in combination with manual tracing, the correlogram can be used for semimanual F0 extraction. If so, F0 extraction can be performed on many voices that cause problems for conventional F0 extractors. To demonstrate the properties of the method it is applied to synthetic and natural voices, among them six pathological voices, which are characterized by roughness, vocal fry, gratings/scrape, hypofunctional breathiness and voice breaks, or combinations of these.

A computer system for acoustic analysis of pathological voices and laryngeal diseases screening

A system for acoustic analysis of pathological voices is proposed. The vocalized part of the voice signal is separated and all glottal cycles are traced by means of a cross-correlation detector. Based on the so determined beginning and duration of all glottal cycles, shimmer, jitter, several harmonics-to-noise ratios and other widely used acoustic parameters are calculated. New parameters are also introduced for estimation of the turbulent noise in voice signals (Turbulent Noise Index — TNI) and for the “breathy” voice characterization (Normalized First Harmonic Energy — NFHE). These parameters are applied for laryngeal pathology detection. An identification accuracy of 96.1% by the K-nearest neighbors’ method has been achieved. The system is clinically tested.

Acoustic analysis of pathological voices. A voice analysis system for the screening of laryngeal diseases.

The authors propose a software system for pathological voice analysis and screening of laryngeal diseases that has the following properties: a) The analysis is totally noninvasive-the microphone is about 30 cm from the patient's mouth, and contact microphones or laryngophones are not used. b) The system is built around a low-cost personal computer (PC). The additional hardware consists of a standard sound card such as "Sound Blaster" (Creative Technology Inc., Paris, France). The card also includes a microphone. Any other analog-to-digital converter board allowing a sampling rate higher than 16 kHz, with 16 bit resolution (Sound Blaster and OROS have been already tested), and a linear phase condenser or electret microphone can be used to minimize the distortions during capture of the signal. c) The software is graphics-driven and user-friendly, therefore no special training is needed in using the system. d) The system is electrically safe since there is no electrical link between patient and PC.

An acoustic analysis of pathological voice and its application to the evaluation of laryngeal pathology

The paper investigates the acoustic measures of pitch perturbations and vocal noise in pathological voice signals in terms of their ability to discriminate between normal and pathological voice status. A brief review of the related work done by other researchers in Japan is also given.

Study on Perceptive and Acoustical Olassification of Pathologic Voices

Analysis of pathological voices seems to be a useful method for the differential diagnosis of various laryngeal diseases. Possible analytic methods are twofold: one is perceptive and the other is acoustical.On 30 voice samples from patients with laryngeal cancers (10), unilateral recurrent nerve paralyses (10), and vocal polyps (10), the author carried out two analytic studies, separately, and compared both results.For the perceptive study, the semantic differential technique was employed as a psychometic procedure with the use of ten pair of polar-opposite adjectives, while for the acoustical study 12 parameters including fundamental frequency, noise components and irregularity of the wave forms of the voices were utilized.Both series of the data were analysed by the Lawley's maximum likelihood method and canonical analysis.The perceptive study showed that there were only two dominant factors, related to the evaluative nature, and loudness sensation of the voices, respectively, which were sufficient in order to classify the three kinds of diseases.In the acoustical study also, two factors were sufficient to classify three kinds of diseases, although identification was rather difficult.The comparison of the results of the perceptive and the acoustical studies revealed that the former was more effective in classifying the laryngeal diseases, as far as the procedures used in this study were concerned.