Modal and Nonmodal Voice Quality Classification Using Acoustic and Electroglottographic Features

Borský, Michal; Mehta, Daryush D.; Stan, Jarrad H. Van; Guðnason, Jón

doi:10.1109/taslp.2017.2759002

Cited by 26 publications

(24 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Therefore, diverse acoustic feature–type groups were explored (e.g., glottal, prosodic, spectral) using the COVAREP speech toolkit [ 32 ]. Previously, COVAREP features have been utilized to investigate and automatically recognize voice quality [ 33 , 34 ], respiratory [ 35 ], voice [ 36 , 37 ], and psychogenic disorders [ 38 , 39 ]. The COVAREP feature set includes 73 individual glottal, prosodic, and spectral features.…”

Section: Methodsmentioning

confidence: 99%

Automatic Detection of COVID-19 Based on Short-Duration Acoustic Smartphone Speech Analysis

Stasak

Huang

Razavi

et al. 2021

J Healthc Inform Res

View full text Add to dashboard Cite

Currently, there is an increasing global need for COVID-19 screening to help reduce the rate of infection and at-risk patient workload at hospitals. Smartphone-based screening for COVID-19 along with other respiratory illnesses offers excellent potential due to its rapid-rollout remote platform, user convenience, symptom tracking, comparatively low cost, and prompt result processing timeframe. In particular, speech-based analysis embedded in smartphone app technology can measure physiological effects relevant to COVID-19 screening that are not yet digitally available at scale in the healthcare field. Using a selection of the Sonde Health COVID-19 2020 dataset, this study examines the speech of COVID-19-negative participants exhibiting mild and moderate COVID-19-like symptoms as well as that of COVID-19-positive participants with mild to moderate symptoms. Our study investigates the classification potential of acoustic features (e.g., glottal, prosodic, spectral) from short-duration speech segments (e.g., held vowel, pataka phrase, nasal phrase) for automatic COVID-19 classification using machine learning. Experimental results indicate that certain feature-task combinations can produce COVID-19 classification accuracy of up to 80% as compared with using the all-acoustic feature baseline (68%). Further, with brute-forced n -best feature selection and speech task fusion, automatic COVID-19 classification accuracy of upwards of 82–86% was achieved, depending on whether the COVID-19-negative participant had mild or moderate COVID-19-like symptom severity.

show abstract

Section: Methodsmentioning

confidence: 99%

Automatic Detection of COVID-19 Based on Short-Duration Acoustic Smartphone Speech Analysis

Stasak

Huang

Razavi

et al. 2021

J Healthc Inform Res

View full text Add to dashboard Cite

show abstract

“…Objective methods involve the processing of dynamic signals obtained from sensors and the calculation of metrics using a vast array of signal processing schemes. The microphone has primarily been used in a vast majority of voice quality studies [10][11][12][13]. Although microphones are convenient, easy to use in the field and have a large bandwidth, the captured voice signals are often contaminated by background noise, and they are distorted by reverberation in the surrounding spaces.…”

Section: Introductionmentioning

confidence: 99%

“…Jitter and shimmer measures of fundamental frequency and loudness perturbations were found to correlate with voice quality based on microphone data [25]. MFCCs and Cepstral Peak Prominence (CPP) have been used to discriminate between modal, breathy, strained and other voice types based on microphone data [11,26]. The difference between the amplitudes of the first two harmonic components on microphone and NSA spectra, H1-H2, was found to be correlated with perceived speech breathiness [13,14].…”

Section: Introductionmentioning

confidence: 99%

“…The Hidden Markov Model (HMM), Support Vector Machine (SVM) and Artificial Neural Network (NN) were used to improve the performance on classification between different voice qualities based on microphone features [29,30]. A Gaussian Mixture Model (GMM) using Mel-Frequency spectral Coefficients (MFCCs) extracted from neck-surface vibration signals achieved 80.2% and 89.5% accuracies in classifying modal, breathy, pressed and rough types at the frame level and utterance level, respectively [11]. Machine learning algorithms critically require, for accurate results, that the calibration be done against a "ground truth", or gold standard.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Discrimination between Modal, Breathy and Pressed Voice for Single Vowels Using Neck-Surface Vibration Signals

et al. 2019

View full text Add to dashboard Cite

The purpose of this study was to investigate the feasibility of using neck-surface acceleration signals to discriminate between modal, breathy and pressed voice. Voice data for five English single vowels were collected from 31 female native Canadian English speakers using a portable Neck Surface Accelerometer (NSA) and a condenser microphone. Firstly, auditory-perceptual ratings were conducted by five clinically-certificated Speech Language Pathologists (SLPs) to categorize voice type using the audio recordings. Intra- and inter-rater analyses were used to determine the SLPs’ reliability for the perceptual categorization task. Mixed-type samples were screened out, and congruent samples were kept for the subsequent classification task. Secondly, features such as spectral harmonics, jitter, shimmer and spectral entropy were extracted from the NSA data. Supervised learning algorithms were used to map feature vectors to voice type categories. A feature wrapper strategy was used to evaluate the contribution of each feature or feature combinations to the classification between different voice types. The results showed that the highest classification accuracy on a full set was 82.5%. The breathy voice classification accuracy was notably greater (approximately 12%) than those of the other two voice types. Shimmer and spectral entropy were the best correlated metrics for the classification accuracy.

show abstract

“…Hence, alternative features for the analysis and classification of phonation types are needed. Mel-frequency cepstral coefficients (MFCCs) derived from speech signals were investigated in [29,34] for classification of phonation types in speech. Similarly the authors of [35] proposed cepstral features derived from high-resolution spectrum obtained by the zerotime windowing (ZTW) method.…”

Section: Introductionmentioning

confidence: 99%

Mel-Frequency Cepstral Coefficients of Voice Source Waveforms for Classification of Phonation Types in Speech

Kadiri¹,

Alku²

2019

Interspeech 2019

View full text Add to dashboard Cite

Voice source characteristics in different phonation types vary due to the tension of laryngeal muscles along with the respiratory effort. This study investigates the use of mel-frequency cepstral coefficients (MFCCs) derived from voice source waveforms for classification of phonation types in speech. The cepstral coefficients are computed using two source waveforms: (1) glottal flow waveforms estimated by the quasi-closed phase (QCP) glottal inverse filtering method and (2) approximate voice source waveforms obtained using the zero frequency filtering (ZFF) method. QCP estimates voice source waveforms based on the source-filter decomposition while ZFF yields source waveforms without explicitly computing the sourcefilter decomposition. Experiments using MFCCs computed from the two source waveforms show improved accuracy in classification of phonation types compared to the existing voice source features and conventional MFCC features. Further, it is observed that the proposed features have complimentary information to the existing features.

show abstract

Modal and Nonmodal Voice Quality Classification Using Acoustic and Electroglottographic Features

Cited by 26 publications

References 44 publications

Automatic Detection of COVID-19 Based on Short-Duration Acoustic Smartphone Speech Analysis

Automatic Detection of COVID-19 Based on Short-Duration Acoustic Smartphone Speech Analysis

Discrimination between Modal, Breathy and Pressed Voice for Single Vowels Using Neck-Surface Vibration Signals

Mel-Frequency Cepstral Coefficients of Voice Source Waveforms for Classification of Phonation Types in Speech

Contact Info

Product

Resources

About