Characteristics Of Speech Signal Research Articles

Decoding Stress with Computer Vision-Based Approach Using Audio Signals for Psychological Event Identification during COVID-19

Interpreting psychological events can be costly and quite complex. It is simple to translate such experiences into a person's spoken and nonverbal cues. The suggested model investigates a computer vision-based method for using an individual's audio signal to identify stressful psychological events. Different people's input speech signals are recorded and compared to the common questionnaire. A series of inquiries pertaining to the second stage of COVID-19 events are included in the questionnaire set. Through additional processing, these speech signals are converted into frequency components by means of the Fast Fourier transformation (FFT) method. A long short-term memory module processes each frequency component and produces temporal information from each frequency band. The features of speech signals are extracted into the temporal frames by this module. The VGG 16 algorithm is used to further classify each temporal frame into stress and un-stress classes. A classifier with 16 layers of architecture is called VGG 16. A feed-forward convolutional neural network called VGG 16 is used to divide the vast array of speech signal features into classes: stressed and unstressed. The proposed model attempts to recognize speech signals as stress indicators. A standard set of questionnaires with a series of interrogation-style questions has been used to develop the stress symptoms in an individual's mind. The audio signals generated by each person's responses are recorded and subsequently analyzed for stress and un-stress classes. The proposed model was able to identify stress in speech signals with 98% accuracy. The time and cost implications of the suggested model are relevant. Medical research is typically costly and time-consuming.LSTM; VGG 16; CNN model; data preprocessing; speech signal.

Linguistic based emotion analysis using softmax over time attention mechanism.

Recognizing the real emotion of humans is considered the most essential task for any customer feedback or medical applications. There are many methods available to recognize the type of emotion from speech signal by extracting frequency, pitch, and other dominant features. These features are used to train various models to auto-detect various human emotions. We cannot completely rely on the features of speech signals to detect the emotion, for instance, a customer is angry but still, he is speaking at a low voice (frequency components) which will eventually lead to wrong predictions. Even a video-based emotion detection system can be fooled by false facial expressions for various emotions. To rectify this issue, we need to make a parallel model that will train on textual data and make predictions based on the words present in the text. The model will then classify the type of emotions using more comprehensive information, thus making it a more robust model. To address this issue, we have tested four text-based classification models to classify the emotions of a customer. We examined the text-based models and compared their results which showed that the modified Encoder decoder model with attention mechanism trained on textual data achieved an accuracy of 93.5%. This research highlights the pressing need for more robust emotion recognition systems and underscores the potential of transfer models with attention mechanisms to significantly improve feedback management processes and the medical applications.

Exploring the Potential of Deep Learning in the Classification and Early Detection of Parkinson's Disease

INTRODUCTION: Parkinson's Disease (PD) is a progressive neurological disorder affecting a significant portion of the global population, leading to profound impacts on daily life and imposing substantial burdens on healthcare systems. Early identification and precise classification are crucial for effectively managing this disease. This research investigates the potential of deep learning techniques in facilitating early recognition and accurate classification of PD.
 OBJECTIVES: The primary objective of this study is to leverage advanced deep learning techniques for the early detection and precise classification of Parkinson's Disease. By utilizing a rich dataset comprising speech signal features extracted from 3000 PD patients, including Time Frequency Features, Mel Frequency Cepstral Coefficients (MFCCs), Wavelet Transform based Features, Vocal Fold Features, and TWQT features, this research aims to evaluate the performance of various deep learning models in PD classification.
 METHODS: The dataset containing diverse speech signal features from PD patients' recordings serves as the foundation for training and evaluating five different deep learning models: ResNet50, VGG16, Inception v2, AlexNet, and VGG19. Each model undergoes training and assessment to determine its capability in accurately classifying PD patients. Performance metrics such as accuracy are employed to evaluate the models' effectiveness.
 RESULTS: The results demonstrate promising potential, with overall accuracies ranging from 89% to 95% across the different deep learning models. Notably, AlexNet emerges as the top-performing model, achieving an accuracy of 95% and demonstrating balanced performance in accurately identifying both true and false PD cases.
 CONCLUSION: This research highlights the significant potential of deep learning in facilitating the early detection and classification of Parkinson's Disease. Leveraging speech signal features offers a non-invasive and cost-effective approach to PD assessment. The findings contribute to the growing body of evidence supporting the integration of artificial intelligence in healthcare, particularly in the realm of neurodegenerative disorders. Further exploration into the application of deep learning in this domain holds promise for advancing PD diagnosis and management.

Research on the Application of Speech Database based on Emotional Feature Extraction in International Chinese Education and Teaching

The advanced analysis of the relationship between acoustic and emotional characteristics of speech signals can effectively improve the interactivity and intelligence of computers. Given the current status of speech recognition and the problems encountered in international Chinese education, the study proposes to extract emotional characteristics to achieve speech construction of the database. Based on considering the emotional characteristics of speech, a hybrid algorithm based on spectral sequence context features is proposed. The DBN-BP algorithm is used to process emotional data of different dimensions, and a speech database is constructed. After testing and analyzing the algorithm model, it is found that the dynamic recognition accuracy of the DBN-BP model fused with emotional features is over 90%, and the negative emotion recognition rates in the three databases are all above 60%. At the same time, the accuracy rate of the model in the algorithm comparison experiment remains above 85%, the data information extraction is relatively complete, and the average test time of less than 1s is less than 3%. The speech database based on multi-emotional feature extraction can effectively provide a new reference for the improvement of the quality of Chinese international education and the improvement of the speech recognition system.

Innovative Teaching of College Vocal Music Courses Aided by Information Integration Technology

Abstract The article adopts the correlation function method to extract the fundamental frequency features of vocal speech signals and processes this information through the cepstrum method. Then, the audio signals are corrected by linear smoothing to match the features of the students’ audio signals with the features of the musical score to determine the accuracy of the vocal music. The study also includes quantization, frame-splitting, and windowing preprocessing of the audio signals, and constructing an unwritten audio signal feature set based on correlation feature selection. The article also proposes an innovative metaphorical method for improving the teaching mode of vocal music courses in colleges and universities. It analyzes the auxiliary role of information technology in teaching vocal music courses by combining simulation experiments with the effects of practical application. The study results show that the students’ vocal singing fundamental frequency range is close to the standard, roughly 123 to 200, with a maximum error value of 24. With the assistance of information fusion technology, the average score increment of the students’ vocal singing reaches 6.088 points, proving this method’s effectiveness in improving the quality of verbal music teaching.

Integration of MFCC Extraction and LSTM Algorithm on PYNQ-Z2 for Enhanced Audio Analysis

The need for Speech Emotion Recognition (SER) is growing since researchers have found it difficult to interpret human emotions from speech data. SER is very interesting yet very challenging task of human-computer interaction (HCI). The SER application can be benefitted depending on the type of feature extraction technique and model used for classification. Deep Learning has made a great impact in the field of audio, image, video, EEG and ECG classification. The speech signal characteristics and classification model affect how well the SER application performs. The paper briefs about deploying Deep Learning Algorithm on FPGA based board i.e., PYNQ-Z2. MFCC feature extraction technique and LSTM model used for classification of human emotion is implemented on the board. Emotion can be predicted using led buttons on the board.

Research on Language Evolution and Language Diversity Based on Chinese Speech Pitch Deviation Features

Abstract This paper constructs a Chinese speech evaluation system based on the characteristics of the vocal mechanism of Chinese speech and uses the detection framework of statistical speech recognition to realize the automatic detection function of pronunciation bias. Assuming that the amplitude spectra of speech and noise obey the generalized Gamma distribution and Gaussian distribution, respectively, the MMSE estimator of the logarithmic spectrum of the speech signal is derived. The speech presence probability under the generalized Gamma speech model is derived as a correction to the MMSE estimation based on language evolution and diversity features. The correct recognition rate of the proposed algorithm reaches 64%, which indicates that the proposed algorithm can effectively suppress the interference of convolutional noise on the speech signal features and improve the subjective perceptual quality of diverse languages.

A method for extracting the semantic features of speech signal recognition based on empirical wavelet transform

The subject of this study is methods for improving the efficiency of semantic coding of speech signals. The purpose of this study is to develop a method for improving the efficiency of semantic coding of speech signals. Coding efficiency refers to the reduction of the information transmission rate with a given probability of error-free recognition of semantic features of speech signals, which will significantly reduce the required source bandwidth, thereby increasing the communication channel bandwidth. To achieve this goal, it is necessary to solve the following scientific tasks: (1) to investigate a known method for improving the efficiency of semantic coding of speech signals based on mel-frequency cepstral coefficients; (2) to substantiate the effectiveness of using the adaptive empirical wavelet transform in the tasks of multiple-scale analysis and semantic coding of speech signals; (3) to develop a method of semantic coding of speech signals based on adaptive empirical wavelet transform with further application of Hilbert spectral analysis and optimal thresholding; and (4) to perform an objective quantitative assessment of the increase in the efficiency of the developed method of semantic coding of speech signals in contrast to the existing method. The following scientific results were obtained during the study: a method of semantic coding of speech signals based on empirical wavelet transform is developed for the first time, which differs from existing methods by constructing a set of adaptive bandpass Meyer wavelet filters with further application of Hilbert spectral analysis to find the instantaneous amplitudes and frequencies of the functions of internal empirical modes, which will allow the identification of semantic features of speech signals and increase the efficiency of their coding; for the first time, it is proposed to use the method of adaptive empirical wavelet transform in the tasks of multiple-scale analysis and semantic coding of speech signals, which will increase the efficiency of spectral analysis by decomposing the high-frequency speech oscillation into its low-frequency components, namely internal empirical modes; the method of semantic coding of speech signals based on mel-frequency cepstral coefficients was further developed, but using the basic principles of adaptive spectral analysis with the help of empirical wavelet transform, which increases the efficiency of this method. Conclusions: We developed a method for semantic coding of speech signals based on empirical wavelet transform, which reduces the encoding rate from 320 to 192 bps and the required bandwidth from 40 to 24 Hz with a probability of error-free recognition of approximately 0.96 (96%) and a signal-to-noise ratio of 48 dB, according to which its efficiency is increased by 1.6 times as compared to the existing method. We developed an algorithm for semantic coding of speech signals based on empirical wavelet transform and its software implementation in the MATLAB R2022b programing language.

Advancements in Speaker Recognition: Exploring Mel Frequency Cepstral Coefficients (MFCC) for Enhanced Performance in Speaker Recognition

Abstract: Speaker recognition, a fundamental capability of software or hardware systems, involves receiving speech signals, identifying the speaker present in the speech signal, and subsequently recognizing the speaker for future interactions. This process emulates the cognitive task performed by the human brain. At its core, speaker recognition begins with speech as the input to the system. Various techniques have been developed for speech recognition, including Mel frequency cepstral coefficients (MFCC), Linear Prediction Coefficients (LPC), Linear Prediction Cepstral coefficients (LPCC), Line Spectral Frequencies (LSF), Discrete Wavelet Transform (DWT), and Perceptual Linear Prediction (PLP). Although LPC and several other techniques have been explored, they are often deemed impractical for real-time applications. In contrast, MFCC stands out as one of the most prominent and widely used techniques for speaker recognition. The utilization of cepstrum allows for the computation of resemblance between two cepstral feature vectors, making it an effective tool in this domain. In comparison to LPC-derived cepstrum features, the use of MFCC features has demonstrated superior performance in metrics such as False Acceptance Rate (FAR) and False Rejection Rate (FRR) for speaker recognition systems. MFCCs leverage the human ear's critical bandwidth fluctuations with respect to frequency. To capture phonetically important characteristics of speech signals, filters are linearly separated at low frequencies and logarithmically separated at high frequencies. This design choice is central to the effectiveness of the MFCC technique. The primary objective of the proposed work is to devise efficient techniques that extract pertinent information related to the speaker, thereby enhancing the overall performance of the speaker recognition system. By optimizing feature extraction methods, this research aims to contribute to the advancement of speaker recognition technology.

Investigation of self-supervised pre-trained models for classification of voice quality from speech and neck surface accelerometer signals

Prior studies in the automatic classification of voice quality have mainly studied the use of the acoustic speech signal as input. Recently, a few studies have been carried out by jointly using both speech and neck surface accelerometer (NSA) signals as inputs, and by extracting mel-frequency cepstral coefficients (MFCCs) and glottal source features. This study examines simultaneously-recorded speech and NSA signals in the classification of voice quality (breathy, modal, and pressed) using features derived from three self-supervised pre-trained models (wav2vec2-BASE, wav2vec2-LARGE, and HuBERT) and using a support vector machine (SVM) as well as convolutional neural networks (CNNs) as classifiers. Furthermore, the effectiveness of the pre-trained models is compared in feature extraction between glottal source waveforms and raw signal waveforms for both speech and NSA inputs. Using two signal processing methods (quasi-closed phase (QCP) glottal inverse filtering and zero frequency filtering (ZFF)), glottal source waveforms are estimated from both speech and NSA signals. The study has three main goals: (1) to study whether features derived from pre-trained models improve classification accuracy compared to conventional features (spectrogram, mel-spectrogram, MFCCs, i-vector, and x-vector), (2) to investigate which of the two modalities (speech vs. NSA) is more effective as input in the classification task with pre-trained model-based features, and (3) to evaluate whether the deep learning-based CNN classifier can enhance the classification accuracy in comparison to the SVM classifier. The results revealed that the use of the NSA input showed better classification performance compared to the speech signal. Between the features, the pre-trained model-based features showed better classification accuracies, both for speech and NSA inputs compared to the conventional features. The two classifiers performed equally well for all the pre-trained model-based features for both speech and NSA signals. It was also found that the HuBERT features performed better than the wav2vec2-BASE and wav2vec2-LARGE features for both speech and NSA inputs. In particular, when compared to the conventional features, the HuBERT features showed an absolute accuracy improvement of 3%–6% for speech and NSA signals in the classification of voice quality.

Performance Improvement of Speech Emotion Recognition Systems by Combining 1D CNN and LSTM with Data Augmentation

In recent years, the increasing popularity of smart mobile devices has made the interaction between devices and users, particularly through voice interaction, more crucial. By enabling smart devices to better understand users’ emotional states through voice data, it becomes possible to provide more personalized services. This paper proposes a novel machine learning model for speech emotion recognition called CLDNN, which combines convolutional neural networks (CNN), long short-term memory neural networks (LSTM), and deep neural networks (DNN). To design a system that closely resembles the human auditory system in recognizing audio signals, this article uses the Mel-frequency cepstral coefficients (MFCCs) of audio data as the input of the machine learning model. First, the MFCCs of the voice signal are extracted as the input of the model. Local feature learning blocks (LFLBs) composed of one-dimensional CNNs are employed to calculate the feature values of the data. As audio signals are time-series data, the resulting feature values from LFLBs are then fed into the LSTM layer to enhance learning on the time-series level. Finally, fully connected layers are used for classification and prediction. The experimental evaluation of the proposed model utilizes three databases: RAVDESS, EMO-DB, and IEMOCAP. The results demonstrate that the LSTM model effectively models the features extracted from the 1D CNN due to the time-series characteristics of speech signals. Additionally, the data augmentation method applied in this paper proves beneficial in improving the recognition accuracy and stability of the systems for different databases. Furthermore, according to the experimental results, the proposed system achieves superior recognition rates compared to related research in speech emotion recognition.

SDTF-Net: Static and dynamic time–frequency network for Speech Emotion Recognition

Speech Emotion Recognition (SER) is an important but challenging research topic in Human–Computer interaction (HCI). As a crucial component influencing the model’s performance, in SER, feature extraction needs to comprehensively consider the static local states, the dynamic change tendency, and the time–frequency characteristics of speech signals to generate meaningful emotional representations and further improve the SER performance. To address this aim, we propose the Static and Dynamic Time–frequency Network called SDTF-Net. To reduce the interference between different features, in SDTF-Net, we introduce two parallel modules, i.e., Time–Frequency Convolutional Neural Network (TF-CNN) and Time–Frequency Transformer (TF-Trans), to generate the static and dynamic emotional features based on the time–frequency information, respectively. Specifically, in the TF-Trans module, we design a three-step strategy to learn the relatively complicated dynamic feature, comprising two parallel Bi-direction Multi-scale Dynamic Activation (BMDA) blocks, two parallel transformer (Trans) blocks, and a time–frequency fusion (TF-Cross) block. Subsequently, we propose an SD-Cross module to effectively fuse the learned static and dynamic features, generating the final meaningful emotional representations. Given the problem of data sparsity, moreover, we employ the data augmentation method called SpecAugment to improve the learning ability of SDTF-Net. From the experimental results on IEMOCAP and MSP-IMPROV datasets, we observe that SDTF-Net outperforms the state-of-the-art methods. Furthermore, extensive ablation experiments further indicate the rationality of the SDTF-Net structure and the necessity of considering dynamic and static features based on time–frequency information comprehensively in SER.

Low-Power Feature-Attention Chinese Keyword Spotting Framework with Distillation Learning

In this paper, we propose a novel Low-Power Feature-Attention Chinese Keyword Spotting Framework based on a depthwise separable convolution neural network (DSCNN) with distillation learning to recognize speech signals of Chinese wake-up words. The framework consists of a low-power feature-attention acoustic model and its learning methods. Different from the existing model, the proposed acoustic model based on connectionist temporal classification (CTC) focuses on the reduction of power consumption by reducing model network parameters and multiply-accumulate (MAC) operations through our designed feature-attention network and DSCNN. In particular, the feature-attention network is specially designed to extract effective syllable features from a large number of MFCC features. This could refine MFCC features by selectively focusing on important speech signal features and removing invalid speech signal features to reduce the number of speech signal features, which helps to significantly reduce the parameters and MAC operations of the whole acoustic model. Moreover, DSCNN with fewer parameters and MAC operations compared with traditional convolution neural networks is adopted to extract effective high-dimensional features from syllable features. Furthermore, we apply a distillation learning algorithm to efficiently train the proposed low-power acoustic model by utilizing the knowledge of the trained large acoustic model. Experimental results thoroughly verify the effectiveness of our model and show that the proposed acoustic model still has better accuracy than other acoustic models with the lowest power consumption and smallest latency measured by NVIDIA JETSON TX2. It has only 14.524 KB parameters and consumes only 0.141 J energy per query and 17.9 ms latency on the platform, which is hardware-friendly.

Central Kurdish Automatic Speech Recognition using Deep Learning

Automatic Speech Recognition (ASR) as an interesting field of speech processing, is nowadays utilized in real applications which are implemented using various techniques. Amongst them, the artificial neural network is the most popular one. Increasing the performance and making these systems robust to noise are among the current challenges. This paper addresses the development of an ASR system for the Central Kurdish language (CKB) using a transfer learning of Deep Neural Networks (DNN). The combination of Mel-Frequency Cepstral Coefficients (MFCCs) for extracting features of speech signals, Long Short-Term Memory (LSTM) with Connectionist Temporal Classification (CTC) output layer is used to create an Acoustic Model (AM) on the AsoSoft CKB speech dataset. Also, we have used the N-gram language model on the collected large text dataset which includes about 300 million tokens. The text corpus is also used to extract a dynamic lexicon model that contains over 2.5 million CKB words. The obtained results show that the use of a DNN improves the results compared to classical statistics modules. The proposed method achieves a 0.22%-word error rate by combining transfer learning and language model adaptation. This result is superior to the best-reported result for the CKB.

End-to-End Deep Convolutional Recurrent Models for Noise Robust Waveform Speech Enhancement.

Because of their simple design structure, end-to-end deep learning (E2E-DL) models have gained a lot of attention for speech enhancement. A number of DL models have achieved excellent results in eliminating the background noise and enhancing the quality as well as the intelligibility of noisy speech. Designing resource-efficient and compact models during real-time processing is still a key challenge. In order to enhance the accomplishment of E2E models, the sequential and local characteristics of speech signal should be efficiently taken into consideration while modeling. In this paper, we present resource-efficient and compact neural models for end-to-end noise-robust waveform-based speech enhancement. Combining the Convolutional Encode-Decoder (CED) and Recurrent Neural Networks (RNNs) in the Convolutional Recurrent Network (CRN) framework, we have aimed at different speech enhancement systems. Different noise types and speakers are used to train and test the proposed models. With LibriSpeech and the DEMAND dataset, the experiments show that the proposed models lead to improved quality and intelligibility with fewer trainable parameters, notably reduced model complexity, and inference time than existing recurrent and convolutional models. The quality and intelligibility are improved by 31.61% and 17.18% over the noisy speech. We further performed cross corpus analysis to demonstrate the generalization of the proposed E2E SE models across different speech datasets.

Learning multi-scale features for speech emotion recognition with connection attention mechanism

Speech emotion recognition (SER) has become a crucial topic in the field of human–computer interactions. Feature representation plays an important role in SER, but there are still many challenges in feature representation such as the inability to predict which features are most effective for SER and the cultural differences in emotion expression. Most previous studies use a single type of feature for the recognition task or conduct early fusion of features. However, a single type of feature cannot well reflect the emotions of speech signals. Also, different features contain different information, direct fusion cannot integrate the advantages of different features. To overcome these challenges, this paper proposes a parallel network for multi-scale SER based on a connection attention mechanism (AMSNet). AMSNet fuses fine-grained frame-level manual features with coarse-grained utterance-level deep features. Meanwhile, it adopts different speech emotion feature extraction modules according to the temporal and spatial features of speech signals, which enriches features and improves feature characterization. The network consists of a frame-level representation learning module (FRLM) based on the time structure and an utterance-level representation learning module (URLM) based on the global structure. Besides, improved attention-based long short-term memory (LSTM) is introduced into FRLM to focus on the frames that contribute more to the final emotion recognition result. In URLM, a convolutional neural network with the squeeze-and-excitation block (SCNN) is introduced to extract deep features. In addition, the connection attention mechanism is proposed for feature fusion, which applies different weights to different features. Extensive experiments are conducted on the IEMOCAP and EmoDB datasets, and the results demonstrate the effectiveness and performance superiority of AMSNet. Our code will be publicly available at https://codeocean.com/capsule/8636967/tree/v1.

A bioinspired configurable cochlea based on memristors.

Cochleas are the basis for biology to process and recognize speech information, emulating which with electronic devices helps us construct high-efficient intelligent voice systems. Memristor provides novel physics for performing neuromorphic engineering beyond complementary metal-oxide-semiconductor technology. This work presents an artificial cochlea based on the shallen-key filter model configured with memristors, in which one filter emulates one channel. We first fabricate a memristor with the TiN/HfOx/TaOx/TiN structure to implement such a cochlea and demonstrate the non-volatile multilevel states through electrical operations. Then, we build the shallen-key filter circuit and experimentally demonstrate the frequency-selection function of cochlea’s five channels, whose central frequency is determined by the memristor’s resistance. To further demonstrate the feasibility of the cochlea for system applications, we use it to extract the speech signal features and then combine it with a convolutional neural network to recognize the Free Spoken Digit Dataset. The recognition accuracy reaches 92% with 64 channels, compatible with the traditional 64 Fourier transform transformation points of mel-frequency cepstral coefficients method with 95% recognition accuracy. This work provides a novel strategy for building cochleas, which has a great potential to conduct configurable, high-parallel, and high-efficient auditory systems for neuromorphic robots.

How Does a Face Mask Impact Speech Perception?

Face masks are mandatory during the COVID-19 pandemic, leading to attenuation of sound energy and loss of visual cues which are important for communication. This study explores how a face mask affects speech performance for individuals with and without hearing loss. Four video recordings (a female speaker with and without a face mask and a male speaker with and without a face mask) were used to examine individuals’ speech performance. The participants completed a listen-and-repeat task while watching four types of video recordings. Acoustic characteristics of speech signals based on mask type (no mask, surgical, and N95) were also examined. The availability of visual cues was beneficial for speech understanding—both groups showed significant improvements in speech perception when they were able to see the speaker without the mask. However, when the speakers were wearing the mask, no statistical significance was observed between no visual cues and visual cues conditions. Findings of the study demonstrate that provision of visual cues is beneficial for speech perception for individuals with normal hearing and hearing impairment. This study adds value to the importance of the use of communication strategies during the pandemic where visual information is lost due to the face mask.

Auditory Adaptation to Speech Signal Characteristics

Auditory Adaptation to Speech Signal Characteristics

Long-term Average Spectrum and Nasal Accelerometry in Sentences of Differing Nasality and Forward-Focused Vowel Productions Under Altered Auditory Feedback

To investigate whether voice focus adjustments can alter the audio-vocal feedback and consequently modulate speech/voice motor control. Speaking with a forward-focused voice was expected to enhance audio-vocal feedback and thus decrease the variability of vocal fundamental frequency (F0). Twenty-two healthy, untrained adults (10 males and 12 females) were requested to sustain vowel /a/ with their natural focus and a forward focus and to naturally read the nasal, oral, and mixed oral-nasal sentences in normal noise-masked auditory conditions. Meanwhile, a miniature accelerometer was externally attached on the noise to detect the nasal vibrations during vocalization. Audio recordings were made and analyzed using the long-term average spectrum (LTAS) and power spectral analysis of F0. Compared with naturally-focused vowel production and oral sentences, forward-focused vowel productions and nasal sentences both showed significant increases in nasal accelerometric amplitude and the spectral power within the range of 200∼300 Hz, and significantly decreased the F0 variability below 3 Hz, which has been reported to be associated with enhanced auditory feedback in our previous research. The auditory masking not only significantly increased the low-frequency F0 variability, but also significantly decreased the ratio of the spectral power within 200∼300 Hz to the power within 300∼1000 Hz for the vowel and sentence productions. Gender differences were found in the correlations between the degree of nasal coupling and F0 stability as well as in the LTAS characteristics in response to noise. Variations in nasal-oral acoustic coupling not only change the formant features of speech signals, but involuntarily influence the auditory feedback control of vocal fold vibrations. Speakers tend to show improved F0 stability in response to a forward-focused voice adjustment.