Analysis and Classification of Cold Speech Using Variational Mode Decomposition

Background/Objectives: Speech produced by individuals with hearing loss differs notably from that of normal-hearing (NH) individuals. Although cochlear implants (CIs) provide sufficient auditory input to support speech acquisition and control, there remains considerable variability in speech intelligibility among CI users. As a result, speech produced by CI talkers often exhibits distinct acoustic characteristics compared to that of NH individuals. Methods: Speech data were obtained from eight cochlear-implant (CI) and eight normal-hearing (NH) talkers, while electroencephalogram (EEG) responses were recorded from 11 NH listeners exposed to the same speech stimuli. Support Vector Machine (SVM) classifiers employing 3-fold cross-validation were evaluated using classification accuracy as the performance metric. This study evaluated the efficacy of Support Vector Machine (SVM) algorithms using four kernel functions (Linear, Polynomial, Gaussian, and Radial Basis Function) to classify speech produced by NH and CI talkers. Six acoustic features—Log Energy, Zero-Crossing Rate (ZCR), Pitch, Linear Predictive Coefficients (LPC), Mel-Frequency Cepstral Coefficients (MFCCs), and Perceptual Linear Predictive Cepstral Coefficients (PLP-CC)—were extracted. These same features were also extracted from electroencephalogram (EEG) recordings of NH listeners who were exposed to the speech stimuli. The EEG analysis leveraged the assumption of quasi-stationarity over short time windows. Results: Classification of speech signals using SVMs yielded the highest accuracies of 100% and 94% for the Energy and MFCC features, respectively, using Gaussian and RBF kernels. EEG responses to speech achieved classification accuracies exceeding 70% for ZCR and Pitch features using the same kernels. Other features such as LPC and PLP-CC yielded moderate to low classification performance. Conclusions: The results indicate that both speech-derived and EEG-derived features can effectively differentiate between CI and NH talkers. Among the tested kernels, Gaussian and RBF provided superior performance, particularly when using Energy and MFCC features. These findings support the application of SVMs for multimodal classification in hearing research, with potential applications in improving CI speech processing and auditory rehabilitation.

Aiming at the shortcoming of harmonic detection in power systems, a method based on Variational Modal Decomposition (VMD) is presented in this paper. This method combines VMD with Fast Fourier Transform (FFT) so that VMD can adaptively find the modal decomposition number 'K'. K intrinsic mode functions (IMF) can be obtained by adaptive VMD decomposition of harmonic signals and Teager energy operator (TEO) analysis can be performed. By calculating the three sample points of the signal via TEO, the instantaneous amplitude and instantaneous frequency of one sample point can be acquired, so as to realize the harmonic detection of the power grid signal. Through the simulation of transient signal, the results of adaptive VMD are tested. At the same time, TEO and Hilbert transform (HT) are used to calculate the IMF respectively. Through contrastive analysis of the caculation, the feasibility and accuracy of the method mentioned in the article can be verified.

Variational mode decomposition is a new signal decomposition method, which can process non-linear and non-stationary signals. It can overcome the problems of mode mixing and compensate for the shortcomings in empirical mode decomposition. Permutation entropy is a method which can detect the randomness and kinetic mutation behavior of a time series. It can be considered for use in fault diagnosis. The complexity of wind power generation systems means that the randomness and kinetic mutation behavior of their vibration signals are displayed at different scales. Multi-scale permutation entropy analysis is therefore needed for such vibration signals. This research investigated a method based on variational mode decomposition and permutation entropy for the fault diagnosis of a wind turbine roller bearing. Variational mode decomposition was adopted to decompose the bearing vibration signal into its constituent components. The components containing key fault information were selected for the extraction of their permutation entropy. This entropy was used as a bearing fault characteristic value. The nearest neighbor algorithm was employed as a classifier to identify faults in a roller bearing. The experimental data showed that the proposed method can be applied to wind turbine roller bearing fault diagnosis.

Incipient fault feature extraction of rolling bearings based on the MVMD and Teager energy operator

Vie des entreprises

The vast number of information packed into the human speech signal makes analysis a tough undertaking. This intricacy is notably noticeable in tasks like speaker recognition, especially when it comes to gender distinction. In this paper, we address this issue by conducting a thorough examination of the effectiveness of various speech features, namely Pitch, Formant Frequency, MFCC (Mel Frequency Cepstral Coefficients), and Chroma, in the context of gender identification in the Tai Language, which is spoken by the Tai people of Assam. In this study, we use machine learning (SVM, KNN, Decision Tree, Neural Network) to analyze speech features (Pitch, Formant Frequency, MFCC, Chroma) for gender identification in the Tai language spoken by the Tai people of Assam. Our results show that MFCC consistently outperforms other features, delivering the highest accuracy rates across all approaches. This demonstrates MFCC’s ability to extract gender information from Tai Language speech signals, suggesting more accurate gender identification systems. Beyond gender identification, our study extends voice analysis in linguistics and improves the application of spoken language data, allowing for improved communication systems and linguistic insights. In summary, our findings highlight the critical significance of MFCC in gender identification in the Tai language, with ramifications that extend far beyond its local context, promising advances in voice analysis and improving our understanding of language and human communication.

In this paper, a new Cable fault location scheme is proposed for the existing flexible DC Cable fault location methods, in which the fault signal is mixed with noise and the traveling wave speed is affected by line parameters. This scheme is based on the propagation characteristics of fault line mode voltage traveling waves in flexible DC transmission. Firstly, the fault voltage is effectively decomposed using Variational Mode Decomposition (VMD). Then, the decomposed intrinsic mode functions (IMF) are reconstructed to obtain the high-frequency components of the recorded fault information. Finally, the traveling wave head is calibrated using the Teager Energy Operator (TEO). A dual terminal ranging algorithm that is not affected by wave velocity has been proposed, which can effectively avoid the impact of fault traveling wave velocity uncertainty on the ranging results. A dual ended flexible DC power grid model was built in PSCAD/EMTDC, and a large number of simulation results show that the algorithm proposed in this paper can accurately calibrate the fault traveling wave head, has little interference from transition resistance, strong noise resistance, and can obtain high-precision fault location results in different fault locations.

A new denoising algorithm and feature extraction algorithm that combine a new kind of permutation entropy (NPE) and variational mode decomposition (VMD) are put forward in this paper. VMD is a new self-adaptive signal processing algorithm, which is more robust to sampling and noise, and also can overcome the problem of mode mixing in empirical mode decomposition (EMD) and ensemble EMD (EEMD). Permutation entropy (PE), as a nonlinear dynamics parameter, is a powerful tool that can describe the complexity of a time series. NPE, a new version of PE, is interpreted as distance to white noise, which shows a reverse trend to PE and has better stability than PE. In this paper, three kinds of ship-radiated noise (SN) signal are decomposed by VMD algorithm, and a series of intrinsic mode functions (IMF) are obtained. The NPEs of all the IMFs are calculated, the noise IMFs are screened out according to the value of NPE, and the process of denoising can be realized by reconstructing the rest of IMFs. Then the reconstructed SN signal is decomposed by VMD algorithm again, and one IMF containing the most dominant information is chosen to represent the original SN signal. Finally, NPE of the chosen IMF is calculated as a new complexity feature, which constitutes the input of the support vector machine (SVM) for pattern recognition of SN. Compared with the existing denoising algorithms and feature extraction algorithms, the effectiveness of proposed algorithms is validated using the numerical simulation signal and the different kinds of SN signal.

The equivalent filter characteristics of variational mode decomposition (VMD) are fully evaluated when applied to the fractional Gaussian noise (fGn) and the application in separating closely spaced modes of vibration system is performed in this paper. VMD is a newly proposed signal decomposition technique, which nonrecursively decomposes a signal into a given number of subsignals (modes), and each mode is mostly compact around a center pulsation. The filter performance of VMD is largely dependent on the constraint parameter and the initialization of center frequencies. In order to extract the desired modes, criteria for the determination of decomposition parameters are established. The initial center frequencies could be simply determined by prior estimated modal frequencies of the analyzed signal, while the constraint parameter is optimized utilizing a genetic algorithm (GA). A two-degree-of-freedom parametric system is considered to evaluate the capability of VMD in the separation of closely spaced modes. Compared with the noise-assisted versions of empirical mode decomposition (EMD) and wavelet packet transform (WPT), the parameter-optimized VMD can successfully separate the closely spaced modes while recovering the most modal information simultaneously. When introduced to the ground vibration test (GVT) of a horizontal tail, the proposed method successfully extracted the first five oscillation modes and identified the modal parameters accurately.

Detecting vowels in a noisy speech signal is a very challenging task. The problem is further aggravated when the noise exhibits speech-like characteristics, e.g., babble noise. In this work, a novel front-end feature extraction technique exploiting variational mode decomposition (VMD) is proposed to improve the detection of vowels in speech data degraded by speech-like noise. Each short-time analysis frame of speech is first decomposed into a set of variational mode functions (VMFs) using VMD. The logarithmic energy present in each of the VMFs is then used as the front-end features for detecting vowels. A three-class classifier (vowel, non-vowel and silence) with acoustic modeling based on long short-term memory (LSTM) architecture is developed on the TIMIT database using the proposed features as well as mel-frequency cepstral coefficients (MFCC). Using the three-class classifier, frame-level time-alignments for a given speech utterance are obtained to detect the vowel regions. The proposed features result in significantly improved performance under noisy test conditions than the MFCC features. Further, the vowel regions detected using the proposed features are also quite different from those obtained through the MFCC. Exploiting the aforementioned differences, the evidences are combined to further improve the detection accuracy.

In this paper, we are introducing speech database consists of 27 Indian languages for analyzing language specific information present in speech. In the context of Indian languages, systematic analysis of various speech features and classification models in view of automatic language identification has not performed, because of the lack of proper speech corpus covering majority of the Indian languages. With this motivation, we have initiated the task of developing multilingual speech corpus in Indian languages. In this paper spectral features are explored for investigating the presence of language specific information. Melfrequency cepstral coefficients (MFCCs) and linear predictive cepstral coefficients (LPCCs) are used for representing the spectral information. Gaussian mixture models (GMMs) are developed to capture the language specific information present in spectral features. The performance of language identification system is analyzed in view of speaker dependent and independent cases. The recognition performance is observed to be 96% and 45% respectively, for speaker dependent and independent environments.

Obstructive Sleep Apnea Hypopnea Syndrome (OSAHS) is a serious chronic sleep disorder. Snoring is a common and easily observable symptom of OSAHS patients. The purpose of this work is to identify OSAHS patients by analyzing the acoustic characteristics of snoring sounds throughout the entire night. Ten types of acoustic features, such as Mel-frequency cepstral coefficients (MFCC), linear prediction coefficients (LPC) and spectral entropy among others, were extracted from the snoring sounds. A fused feature selection algorithm based on ReliefF and Max-Relevance and Min-Redundancy (mRMR) was proposed for optimal feature set selection. Four types of machine learning models were then applied to validate the effectiveness of OSAHS patient identification. The results show that the proposed feature selection algorithm can effectively select features with high contribution, including MFCC and LPC. Based on the selected top-20 features and using a support vector machine model, the accuracies in identifying OSAHS patients under the thresholds of AHI = 5,15, and 30, were 100%, 100%, and 98.94%, respectively. This indicates that the proposed model can effectively identify OSAHS patients.

The aim of the study was to inspect the acoustic properties and sleep characteristics of a preapneic snoring sound. The feasibility of forecasting upcoming respiratory events by snoring sound was also investigated. Participants with habitual snoring or a heavy breathing sound during sleep were recruited consecutively. Polysomnography was conducted, and snoring-related breathing sound was recorded simultaneously. Acoustic features and sleep features were extracted from 30-second samples, and a machine learning algorithm was used to establish 2 prediction models. A total of 74 eligible participants were included. Model 1, tested by 5-fold cross-validation, achieved an accuracy of 0.92 and an area under the curve of 0.94 for respiratory event prediction. Model 2, with acoustic features and sleep information tested by Leave-One-Out cross-validation, had an accuracy of 0.78 and an area under the curve of 0.80. Sleep position was found to be the most important among all sleep features contributing to the performance of the 2 models. Preapneic sound presented unique acoustic characteristics, and snoring-related breathing sound could be deployed as a real-time apneic event predictor. The models, combined with sleep information, serve as a promising tool for an early warning system to forecast apneic events. Wang B, Yi X, Gao J, etal. Real-time prediction of upcoming respiratory events via machine learning using snoring sound signal. J Clin Sleep Med. 2021;17(9):1777-1784.

Intra-speaker variations due to perceptually induced stress or emotion adversely affect speech recognition system performance. In this paper, we combine auditory based (Mel frequency cepstral coefficients and linear predictive cepstral coefficients) features and articulatory based (voicedness) features for robust speech recognition. Voicedness features are derived using linear and teager energy operator (TEO) based nonlinear fast Fourier transform (FFT) spectra. Nonlinear properties are analyzed in both the time and frequency domains. In addition, we investigate the sensitivity of all these FFT spectra to stress and observe the performance of individual FFT spectra. The system is tested using stressed speech data from the speech under simulated and actual stress (SUSAS) database. The results show that articulatory based features help to improve the system performance. Furthermore, significant performance improvement has been observed when using the FFT spectrum which is less sensitive to stress

Precise fault localization is vital for enhancing power grid reliability, reducing outage duration, and lowering maintenance expenses. This paper presents a fast and effective method that utilizes the application of voltage pulses with opposing polarities and identical durations to the transmission line. Analysis of traveling wave propagation under both normal and fault conditions indicates that the initial voltage wave observed in the non-injected phase stems from reflections at the fault point. The arrival time of the wavefront is accurately determined using variational mode decomposition (VMD) combined with the Teager energy operator (TEO) to pinpoint the fault location. Simulations conducted in PSCAD validate the efficacy of the proposed approach. This method demonstrates advantages including straightforward implementation, swift response, and superior accuracy compared to conventional techniques.