Deep Learning-Based Voice Pathology Detection From Electroglottography

This paper presents a convolutional neural network-based pathological voice detection system using speech and electroglottographic (EGG) signals. Speech signals have been popularly used to detect voice pathology. Recently, the EGG signals have drawn considerable attention from researchers in this field. They argued that the EGG signals could detect the vocal fold vibration more accurately than speech signals and hence can be considered more appropriate for voice pathology detection. This work investigates the effectiveness of the EGG and speech signals in detecting pathological voices using sustained vowel (“/a/”) samples collected from the Saarbrücken Voice Database (SVD). The Mel frequency cepstral coefficients (MFCCs) extracted from the speech and EGG samples are employed as discerning features for this investigation. The results show that the proposed system achieves a higher accuracy (more than 23%) in identifying dysphonic voices from healthy ones with speech signals.

MMHFNet: Multi-modal and multi-layer hybrid fusion network for voice pathology detection

Automatic voice pathology detection and classification systems effectively contribute to the assessment of voice disorders, enabling the early detection of voice pathologies and the diagnosis of the type of pathology from which patients suffer. This paper concentrates on developing an accurate and robust feature extraction for detecting and classifying voice pathologies by investigating different frequency bands using autocorrelation and entropy. We extracted maximum peak values and their corresponding lag values from each frame of a voiced signal by using autocorrelation as features to detect and classify pathological samples. We also extracted the entropy for each frame of the voice signal after we normalized its values to be used as the features. These features were investigated in distinct frequency bands to assess the contribution of each band to the detection and classification processes. Various samples of the sustained vowel /a/ for both normal and pathological voices were extracted from three different databases in English, German, and Arabic. A support vector machine was used as a classifier. We also performed u-Tests to investigate if there is a significant difference between the means of the normal and pathological samples. The best achieved accuracies in both detection and classification varied depending on the used band, method, and database. The most contributive bands in both detection and classification were between 1000 and 8000 Hz. The highest obtained accuracies in the case of detection were 99.69%, 92.79%, and 99.79% for Massachusetts eye and ear infirmary (MEEI), Saarbrücken voice database (SVD), and Arabic voice pathology database (AVPD), respectively. However, the highest achieved accuracies for classification were 99.54%, 99.53%, and 96.02% for MEEI, SVD, and AVPD, correspondingly, using the combined feature.

Deep learning methods are data-driven. But for pathological voice detection, it is difficult to obtain high-quality labeled data. In this work, a weakly supervised learning Method is presented to improve the quality of existing datasets by learning sample weights and fine-grained labels. First, A convolutional neural network (CNN) is devised as the basic architecture to detect the pathological voice. Then, a proposed self-training algorithm is used to iteratively run and automatically learn the sample weights and fine-grained labels. These learned sample weights and fine-grained labels are used to train the CNN model from scratch. The experiment results on the Saarbrucken Voice database show that the diagnosis accuracy improved from 75.7 to 82.5%, with a 6.8% improvement in accuracy over the CNN models trained with the original dataset. This work demonstrates that the weakly supervised learning method can significantly improve the classification performance to distinguish pathological voice and healthy voice.KeywordsWeakly supervised learningPathological voiceAcoustic analysisDeep learningConvolutional neural network

The voice pathology detection using automatic classification systems is a useful way to diagnose voice diseases. In this paper, we propose a novel tool to detect voice pathology based on an incremental possibilistic SVM-HMM method which can be applied to serval practical applications using non-stationary or a very large-scale data in purpose to reduce the memory issues faced during the storage of the kernel matrix. The proposed system includes the steps of using SVM to incrementally compute possibilitic probabilities and then they will be used by HMM in order to detect voice pathologies. We evaluated the proposed method on the task of the detection of voice pathologies using voices samples from the Massachusetts Eye and Ear Infirmary Voice and Speech Laboratory (MEEI) database. According to the detection rates obtained by our system, the performance sounds robust, efficient and speed applied to a task of voices pathology detection.

A novel methodology, based on the estimation of nonlinear dynamics features, is presented for automatic detection of pathologies in the phonatory system considering continuous speech records (text-dependent). The proposed automatic segmentation and characterization of the voice registers does not require the estimation of the pitch period, therefore it doesn't depend on the gender and intonation of the patients. A robust methodology for finding the features that better discriminate between healthy and pathological voices and also for analyzing the affinity among them is also presented. An average success rate of 95% ± 3.54% in the automatic detection of voice pathologies is achieved considering only six features. The results indicate that nonlinear dynamics is a good alternative for automatic detection of abnormal phonations in continuous speech.

LPC Spectrum First Peak Analysis for Voice Pathology Detection

In this paper, a mode decomposition analysis-based adaptive approach is proposed to provide high diagnostic performance for automated voice pathology detection systems. The aim of the study is to develop a reliable and effective system using adaptive cepstral domain features derived from the empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD), and complete empirical mode decomposition with adaptive noise (CEEMDAN) methods. The descriptive feature sets are obtained by applying mel-frequency cepstral coefficients (MFCCs) and their derivatives, linear predictive coefficients (LPCs) and linear predictive cepstral coefficients (LPCCs) techniques to each decomposition level. The class-balanced data are generated on the VOice ICar fEDerico II database samples using the synthetic minority oversampling technique (SMOTE). The ReliefF algorithm is used to select the most effective and distinctive features. A combination of selected features and a support vector machine (SVM) classifier is used to identify pathological voices. In the pathology detection approach, the results show that the cepstral features based on EMD and SVM-cubic achieves the highest performance with 99.85\% accuracy, 99.85\% F1-score and 0.997 Matthews correlation coefficient (MCC). In pathology-type classification, the cepstral features based on EEMD and SVM-quadratic approach provided the highest performance with 96.49\% accuracy, 96.46\% F1 and 0.949 MCC values. The comprehensive results of this study reveal that mode decomposition-based approaches are more successful and effective than traditional methods for detection and classification of pathological voices.

In recent years, voice pathology detection (VPD) has received considerable attention because of the increasing risk of voice problems. Several methods, such as support vector machine and convolutional neural network-based models, achieve good VPD performance. To further improve the performance, we use a self-supervised pretrained model as feature representation instead of explicit speech features. When the pretrained model is fine-tuned for VPD, an overfitting problem occurs due to a domain shift from conversation speech to the VPD task. To mitigate this problem, we propose an adversarial task adaptive pretraining (A-TAPT) approach by incorporating adversarial regularization during the continual learning process. Experiments on VPD using the Saarbrucken Voice Database show that the proposed A-TAPT improves the unweighted average recall (UAR) by an absolute increase of 12.36% and 15.38% compared with SVM and ResNet50, respectively. It is also shown that the proposed A-TAPT achieves a UAR that is 2.77% higher than that of conventional TAPT learning.

Reproducible Machine Learning-Based Voice Pathology Detection: Introducing the Pitch Difference Feature.

A novel voice pathology detection technique through deep neural networks utilizing speech and electroglottographic signals

Evaluation of Voice Pathology Based on the Estimation of Vocal Fold Biomechanical Parameters

Multimodal Voice Pathology Detection With Voice Signal and Integrated Electroglottographic (EGG) Signal From the Saarbrücken Voice Database.

Automatic voice pathology detection and classification using vocal tract area irregularity

Voice recognition systems have become increasingly important in recent years due to the growing need for more efficient and intuitive human-machine interfaces. The use of Hybrid LSTM networks and deep learning has been very successful in improving speech detection systems. The aim of this paper is to develop a novel approach for the detection of voice pathologies using a hybrid deep learning model that combines the Bidirectional Long Short-Term Memory (BiLSTM) and the Convolutional Neural Network (CNN) architectures. The proposed model uses a combination of temporal and spectral features extracted from speech signals to detect the different types of voice pathologies. The performance of the proposed detection model is evaluated on a publicly available dataset of speech signals from individuals with various voice pathologies(MEEI database). The experimental results showed that the hybrid BiLSTM-CNN model outperforms several classifiers by achieving an accuracy of 98.86\%. The proposed model has the potential to assist health care professionals in the accurate diagnosis and treatment of voice pathologies, and improving the quality of life for affected individuals.