The Massachusetts Eye and Ear Infirmary (MEEI) database is an international-standard training database for voice pathology detection (VPD) systems. However, there is a class-imbalanced distribution in normal and pathological voice samples and different types of pathological voice samples in the MEEI database. This study aimed to develop a VPD system that uses the fuzzy clustering synthetic minority oversampling technique algorithm (FC-SMOTE) to automatically detect and classify four types of pathological voices in a multi-class imbalanced database. The proposed FC-SMOTE algorithm processes the initial class-imbalanced dataset. A set of machine learning models was evaluated and validated using the resulting class-balanced dataset as an input. The effectiveness of the VPD system with FC-SMOTE was further verified by an external validation set and another pathological voice database (Saarbruecken Voice Database (SVD)). The experimental results show that, in the multi-classification of pathological voice for the class-imbalanced dataset, the method we propose can significantly improve the diagnostic accuracy. Meanwhile, FC-SMOTE outperforms the traditional imbalanced data oversampling algorithms, and it is preferred for imbalanced voice diagnosis in practical applications.

Currently, sound source localization (SSL) techniques based on deep learning mainly rely on traditional signal processing methods to generate input features. Nevertheless, the applicability of these features in various environments shows significant differences. This study proposes a new single SSL model, called the icosahedral feature attention network (IFAN), to overcome this limitation. The proposed IFAN not only uses steered response power with phase transform (SRP-PHAT) but also develops steered response power with least mean square (SRP-LMS) as inputs of the network. The IFAN network encodes spatial position information into convolution kernels by introducing icosahedral convolutions. In addition, it adaptively learns optimal feature weights based on the input acoustic environment using the sigmoid function to capture the spatial distribution information of the sound source. For single-source SSL and tracking scenarios, the proposed method on the localization and tracking (LOCATA) challenge data corpus outperforms other state-of-the-art models. Moreover, it is capable of learning complementary information even in acoustic simulations involving a wide range of reverberations. The proposed IFAN can thus enhance the robustness and performance in different environments.

Voice production occurs through vocal cord and vibration coupled to glottal airflow. Vocal cord lesions affect the vocal system and lead to voice disorders. In this paper, a pathological voice source analysis system is designed. This study integrates nonlinear dynamics with an optimized asymmetric two-mass model to explore nonlinear characteristics of vocal cord vibration, and changes in acoustic parameters, such as fundamental frequency, caused by distinct subglottal pressure and varying degrees of vocal cord paralysis are analyzed. Various samples of sustained vowel /a/ of normal and pathological voices were extracted from MEEI (Massachusetts Eye and Ear Infirmary) database. A fitting procedure combining genetic particle swarm optimization and a quasi-Newton method was developed to optimize the biomechanical model parameters and match the targeted voice source. Experimental results validate the applicability of the proposed model to reproduce vocal cord vibration with high accuracy, and show that paralyzed vocal cord increases the model coupling stiffness.

Due to the complexity and non-stationarity of the voice generation system, the nonlinearity of speech signals cannot be accurately quantified. Recently, the recurrence quantification analysis method has been used for voice disorder detection. In this paper, multiscale recurrence quantification measures (MRQMs) are proposed. The signals are reconstructed in the high-dimensional phase space at the equivalent rectangular bandwidth scale. Recurrence plots (RPs) combining the characteristics of human auditory perception are drawn with an appropriate recurrence threshold. Based on the above, the nonlinear dynamic recurrence features of the speech signal are quantized from the recurrence plot of each frequency channel. Furthermore, this paper explores the recurrence quantification thresholds that are most suitable for pathological voices. Our results show that the proposed MRQMs with support vector machine (SVM), random forest (RF), Bayesian network (BN) and Local Weighted Learning (LWL) achieve an average accuracy of 99.45%, outperforming traditional features and other complex measurements. In addition, MRQMs also have the potential for multi-classification of voice disorder, achieving an accuracy of 89.05%. This study demonstrates that MRQMs can characterize the recurrence characteristic of pathological voices and effectively detect voice disorders.

Sulcus vocalis is an indentation parallel to the edge of vocal fold, which may extend into the cover and ligament layer of the vocal fold or deeper. The effects of sulcus vocalis depth on phonation and the vocal cord vibrations are investigated in this study. The three-dimensional laryngeal models were established for healthy vocal folds (0 mm) and different types of sulcus vocalis with the typical depth of 1 mm, 2 mm, and 3 mm. These models with fluid-structure interaction (FSI) are computed numerically by sequential coupling method, which includes an immersed boundary method (IBM) for modelling the glottal airflow, a finite-element method (FEM) for modelling vocal fold tissue. The results show that a deeper sulcus vocalis in the cover layer decreases the vibrating frequency of vocal folds and expands the prephonatory glottal half-width which increases the phonation threshold pressure. The larger sulcus vocalis depth makes vocal folds difficult to vibrate and phonate. The effects of sulcus vocalis depth suggest that the feature such as phonation threshold pressure could assist in the detection of healthy vocal folds and different types of sulcus vocalis.

A pathological voice detection system is designed to detect pathological characteristics of vocal cords from speech. Such systems are particularly susceptible to domain mismatch where the training and testing data come from the source and target domains, respectively. Due to the difference in speech disease etiology, recording environment, and device, etc., the feature distributions of source and target domain are quite different. Meanwhile, considering the high costs of annotating labels, it is hard to acquire labeled data in the target domain. This paper attempts to formulate cross-domain pathological voice detection as an unsupervised domain adaptation problem. Joint subspace transfer learning (JSTL) aims to find a projection matrix to transform source and target domain data into a common space. The maximum mean discrepancy function is used to measure the divergence across databases. Intra-class and inter-class distance act as regularization to guarantee the maximum separability between different classes. A graph matrix is constructed to help transfer knowledge from the relevant source data to the target data. Three popular pathological voice databases were selected in this paper. For six cross-database experiments, the accuracy of the method proposed increased by up to 15%. For different voice categories, the category of structural voice showed the most significant increase, nearly 20%.