Voice Utterances Research Articles

Attention-Based Fusion of Ultrashort Voice Utterances and Depth Videos for Multimodal Person Identification

Multimodal deep learning, in the context of biometrics, encounters significant challenges due to the dependence on long speech utterances and RGB images, which are often impractical in certain situations. This paper presents a novel solution addressing these issues by leveraging ultrashort voice utterances and depth videos of the lip for person identification. The proposed method utilizes an amalgamation of residual neural networks to encode depth videos and a Time Delay Neural Network architecture to encode voice signals. In an effort to fuse information from these different modalities, we integrate self-attention and engineer a noise-resistant model that effectively manages diverse types of noise. Through rigorous testing on a benchmark dataset, our approach exhibits superior performance over existing methods, resulting in an average improvement of 10%. This method is notably efficient for scenarios where extended utterances and RGB images are unfeasible or unattainable. Furthermore, its potential extends to various multimodal applications beyond just person identification.

Mining Hidden Partitions of Voice Utterances using Fuzzy Clustering for Generalized Voice Spoofing Countermeasures

Mining Hidden Partitions of Voice Utterances using Fuzzy Clustering for Generalized Voice Spoofing Countermeasures

A Proposed Speaker Recognition Method Based on Long-Term Voice Features and Fuzzy Logic

Speaker recognition depends on specific predefined steps. The most important steps are feature extraction and features matching. In addition, the category of the speaker voice features has an impact on the recognition process. The proposed speaker recognition makes use of biometric (voice) attributes to recognize the identity of the speaker. The long-term features were used such that maximum frequency, pitch and zero crossing rate (ZCR). In features matching step, the fuzzy inner product was used between feature vectors to compute the matching value between a claimed speaker voice utterance and test voice utterances. The experiments implemented using (ELSDSR) data set. These experiments showed that the recognition accuracy is 100% when using text dependent speaker recognition.

Investigation and Evaluation of Glottal Flow Waveform for Voice Pathology Detection

Automatic voice pathology detection can provide objective estimation and prevention in the early stages of voice diseases. Glottal flow waveform directly reflects the state of glottal excitation. Extracting acoustic features from glottal source signals may contribute to the detection of pathological voice. To improve the performance of voice pathology detection, this article investigates the contribution of the glottal flow waveform for pathological voice detection by evaluating the classification result using features extracted from raw speech utterances and corresponding glottal flow waveforms. The individual feature sets used are extracted from raw or glottal voice utterances with identical parameter settings, which are openSMILE acoustic features, audio features computed by Moving Picture Experts Group-7 standard and classical glottal source features. In addition, a feature selection method in terms of the wrapper approach is used to combine the single features ranked by using the Fisher discrimination ratio. Voice pathology detection experiments were carried out using Random Forest. The best accuracies of 88.52% for the Saarbrucken Voice database and 100.00% for the Massachusetts Eye and Ear Infirmary database are achieved using the combined feature set extracted from the glottal source signal, with improvement of 0.44-3.13% in the accuracies obtained by using raw speech utterances. Compared to state-of-the-art methods, the proposed method achieves the highest accuracy for the Massachusetts Eye and Ear Infirmary database and an increase of 2.75-17.16% in detection accuracy compared to other conventional pipeline systems for the Saarbrucken Voice databse. The experimental results demonstrate that using glottal flow waveform as source signal can improve the performance of pathological voice detection.

Gender classification from fused multi-fingerprint types

ABSTRACT There has been a growing need for demographic information in security applications. These range from token-based demographic information elicitation to automatic gender classification or age estimation from biometric traits. Gender has been classified from facial images, voice utterances, fingerprints, hand images and emerging biometrics in the literature using local statistical or structural feature descriptors extracted from these traits. We propose a deep learning based convolutional neural network architecture for classifying gender from fingerprints of each of the five finger types and evaluate performances across trained models. We demonstrate that performance can be improved by classifying gender from fingerprints of fused combinations amongst the five right-hand finger types. The default method has been to classify gender from the index finger. However, our results show that certain finger types classify a certain gender better than the other. Leveraging on these varying strengths of the finger types we employ a fusion scheme at the abstract level, of odd number of models, trained with these fingerprint types to improve performance. Male, female and overall classification accuracies of the best fusion model are 94.7%, 88.0% and 91.3%, thus, proffering 31.02%, 7.82% and 18.72% improvement, respectively.

Glottal Source Information for Pathological Voice Detection

Automatic methods for the detection of pathological voice from healthy speech can be considered as potential clinical tools for medical treatment. This study investigates the effectiveness of glottal source information in the detection of pathological voice by comparing the classical pipeline approach to the end-to-end approach. The traditional pipeline approach consists of a feature extractor and a separate classifier. In the former, two sets of glottal features (computed using the quasi-closed phase glottal inverse filtering method) are used together with the widely used openSMILE features. Using both the glottal and openSMILE features extracted from voice utterances and the corresponding healthy/pathology labels, support vector machine (SVM) classifiers are trained. In building end-to-end systems, both raw speech signals and raw glottal flow waveforms are used to train two deep learning architectures: (1) a combination of convolutional neural network (CNN) and multilayer perceptron (MLP), and (2) a combination of CNN and long short-term memory (LSTM) network. Experiments were carried out using three publicly available databases, including dysarthric (the UA-Speech database and the TORGO database) and dysphonic voices (the UPM database). The performance analysis of the detection system based on the traditional pipeline approach showed best results when the glottal features were combined with the baseline openSMILE features. The results of the end-to-end approach indicated higher accuracies (about 2-3 % improvement in all three databases) when glottal flow was used as the raw time-domain input (87.93 % for UA-Speech, 81.12 % for TORGO and 76.66 % for UPM) compared to using raw speech waveform (85.12 % for UA-Speech, 78.83 % for TORGO and 73.71 % for UPM). The evaluation of both approaches demonstrate that automatic detection of pathological voice from healthy speech benefits from using glottal source information.

Tonal languages speech synthesis using an indirect pitch markers and the quantitative target approximation methods

Synthesizing tones plays an important role in text-to-speech systems of tonal languages. To accomplish this, the two important steps are to determine the pitch markers of voice utterances and synthesize F0 trajectories for lexical tones. In this paper, we propose two efficient algorithms, one of them is to locate the pitch markers at the peaks of the cumulative signal of each voiced part of the input utterance and the other is to generate F0 trajectories of tones with quantitative target approximation (qTA) parameters of Xu model. The experimentation has shown that the proposed algorithms present pitch markers with high accuracy which has enabled us to generate tones with complex shapes.

On an improvement of the old voice by the reflection of the amplified voice

When people get older, they lose their energy and get tired of everything. At this time, the voice utterance ages and gradually changes to bass, and the sounding frequency of the first resonance frequency decreases with age. Voice phonation is produced by the release of pressure from the lungs through the vocal folds, causing resonance in the vocal tract. In this study, we proposed a new method to improve the aging of vocal organs with their voices using the voice utterance characteristics. We have to hear the vocal organs of the experimenter as usual loudness over a cup that contains the sound recorded voice, the voice-aging properties were measured in one-minute increments. In this study, the average sound pressure level when the aging voice removal was restored to + 6.2dB spectrum of the high-frequency range is improved to the average size + 13.7dB. And the experimenter showed regained the confidence of the utterance to think that his voice clarity improvement over the ages.

Time-domain speech enhancement using generative adversarial networks

Speech enhancement improves recorded voice utterances to eliminate noise that might be impeding their intelligibility or compromising their quality. Typical speech enhancement systems are based on regression approaches that subtract noise or predict clean signals. Most of them do not operate directly on waveforms. In this work, we propose a generative approach to regenerate corrupted signals into a clean version by using generative adversarial networks on the raw signal. We also explore several variations of the proposed system, obtaining insights into proper architectural choices for an adversarially trained, convolutional autoencoder applied to speech. We conduct both objective and subjective evaluations to assess the performance of the proposed method. The former helps us choose among variations and better tune hyperparameters, while the latter is used in a listening experiment with 42 subjects, confirming the effectiveness of the approach in the real world. We also demonstrate the applicability of the approach for more generalized speech enhancement, where we have to regenerate voices from whispered signals.

Building LSTM Neural Network Based Speaker Identification System

In this paper, we are analyzing the results of native Lithuanian speaker recognition and identification using long short-term memory deep neural network. We look at recognition accuracy and identify further potential improvements. Dataset used for training and speaker recognition consists of over 370 unique speakers, who provide their voice utterances in Lithuanian language. In this paper we present results that are derived from part of this dataset. DOI: 10.15181/csat.v6i1.1579

Development of Personal Authentication System with Two Elements of Image Selection and Voice Utterance

Development of Personal Authentication System with Two Elements of Image Selection and Voice Utterance

Speech Emotion Classification using Machine Learning

recent years, the interaction between humans and machines has become an issue of concern. This paper results from study of various researches related to the investigation of the six basic human emotions which include anger, dislike, fear, happiness, sadness and surprise. (1, 3) Feature extraction is done from various voice utterances recorded from different persons. The various features like pitch, energy, fundamental frequency are extracted from the utterances using respective feature extraction algorithms. After feature extraction procedure, the extracted features are classified under the basic six emotions using various machine learning algorithms. (1, 3, 4)And using the different algorithms the classification accuracy is measured for each algorithm respectively. Various acoustic and prosodic features are extracted from the speech recorded and then classified under different emotional category using machine learning tools. (7) This paper discusses how feature extraction through speech samples and then classification of the extracted features under different emotions is performed. KeywordsEmotion Recognition Models: Distributed speech recognition model; Feature Extraction; Classification Algorithms.

Analysis of Voice Modifications for Persons After Tonsillectomy

The goal of the research described in the present paper was the determination of modi cation range for voice acoustic parameters resulting from tonsillectomy. Within the scope of the described research program, an attempt has been made to determine the changes of selected voice parameters for persons after such a treatment and also to elaborate work out some premises for prediction of potential voice modi cations for persons who have not yet decided to undergo such a treatment. In order to achieve the goal, analyses have been carried out for voice utterances of persons before the tonsillectomy surgery and after the treatment. The rst voice recordings took place between one and three days before the surgery. The post-treatment recordings have been carried out about 6 weeks after the surgery, as a procedure accompanying the follow-up examinations. In the present paper, an analysis has been carried out concerning phonemes /a/, /e/, /i/, and /u/ with prolonged phonation. The completed research shows that for evaluation of voice modi cation in the aspect of changes resulting from tonsillectomy, the most useful parameters are some of the mel-cepstral coe cients, the formant frequencies, and also the relative power coe cients.

Handing-over motion model based on analyses of hand motions and voice utterances of humans

Handing-over motion model based on analyses of hand motions and voice utterances of humans

CLASSIFICATION OF PARKINSON'S DISEASE PATIENTS USING NONLINEAR PHONETIC FEATURES AND MEL-FREQUENCY CEPSTRAL ANALYSIS

This paper presents a combinational feature extraction approach using voice utterances for discriminating Parkinson's disease (PD) patients from healthy people. The proposed feature set consists of seven nonlinear phonetic features and 13 usual Mel-frequency cepstral coefficients (MFCCs). In this research, two new features — EDC-PIS (energy distribution coefficient of peak index series) and EDC-PMS (energy distribution coefficient of peak magnitude series) — were introduced, which are robust to many uncontrollable confounding effects such as noisy environments. The nonlinear phonetic features comprise recurrent period density entropy (RPDE), detrended fluctuation analysis (DFA), noise-to-harmonic ratio (NHR), fractal dimension (FD), pitch period entropy (PPE), EDC-PIS, and EDC-PMS. MFCC features have been widely used in voice processing tasks and therefore are good candidates to be used for the voice processing of PD subjects. The dataset used was composed of a range of 200 voice utterances from 25 PD subjects with different severity levels, and 10 normal persons. Using voice utterances from healthy and PD subjects, a 20-dimensional final feature set using MFCCs and nonlinear features is composed. Finally, a multilayer perceptron (MLP) neural network classifier with one hidden layer was used to discriminate PD subjects. Also, the proposed system was used for classification of mild and severe PD subjects. We obtained 97.5% overall correct classification performance for the discrimination of PD. In addition, we obtained 95.5% overall accuracy for the discrimination of mild and severe PD subjects.

The eye-voice lead during oral reading in developmental dyslexia

In reading aloud, the eye typically leads over voice position. In the present study, eye movements and voice utterances were simultaneously recorded and tracked during the reading of a meaningful text to evaluate the eye-voice lead in 16 dyslexic and 16 same-age control readers. Dyslexic children were slower than control peers in reading texts. Their slowness was characterized by a great number of silent pauses and sounding-out behaviors and a small lengthening of word articulation times. Regarding eye movements, dyslexic readers made many more eye fixations (and generally smaller rightward saccades) than controls. Eye movements and voice (which were shifted in time because of the eye-voice lead) were synchronized in dyslexic readers as well as controls. As expected, the eye-voice lead was significantly smaller in dyslexic than control readers, confirming early observations by Buswell (1921) and Fairbanks (1937). The eye-voice lead was significantly correlated with several eye movements and voice parameters, particularly number of fixations and silent pauses. The difference in performance between dyslexic and control readers across several eye and voice parameters was expressed by a ratio of about 2. We propose that referring to proportional differences allows for a parsimonious interpretation of the reading deficit in terms of a single deficit in word decoding. The possible source of this deficit may call for visual or phonological mechanisms, including Goswami's temporal sampling framework.

3203 人の手渡し軌道及び発声解析に基づく直交型手渡しロボットシステムの開発

3203 人の手渡し軌道及び発声解析に基づく直交型手渡しロボットシステムの開発

Women’s Voice and Religious Utterances in Ancient Greece

This paper tackles the issue of women and religion through a particular looking glass: religious utterances such as curses, supplication, and prayer, as reflected in some passages from ancient Greek epic and tragedy—pivotal literary genres in the ideological discourse of the Greek polis.

Voice containment device for a telephone

A telephone voice containment device is described. It includes a housing which is shaped to receive a voice pick-up mechanism of a telephone in order to capture the voice utterances to be picked up by such phone, and at the same time prevent the voice utterances from escaping to the area surrounding the telephone. An elastic band is included to hold the telephone casing and, hence, its pick-up mechanism in the housing. A sound absorbing structure is most desirably included within the housing to absorb voice utterances not needed by the telephone pick-up mechanism.

Synthetic speech in foreign language learning: an evaluation by learners

Can synthetic speech be utilized in foreign language learning as natural speech? In this paper, we evaluated synthetic speech from the viewpoint of learners in order to find out an answer. The results pointed out that learners do not recognize remarkable differences between synthetic voices and natural voices for the words with short vowels and long vowels when they try to understand the meanings of the sounds. The data explicates that synthetic voice utterances of sentences are easier to understand and more acceptable by learners compared to synthetic voice utterances of words. In addition, the ratings on both synthetic voices and natural voices strongly depend upon the learners’ listening comprehension abilities. We conclude that some synthetic speech with specific pronunciations of vowels may be suitable for listening materials and suggest that evaluating TTS systems by comparing synthetic speech with natural speech and building a lexical database of synthetic speech that closely approximates natural speech will be helpful for teachers to readily use many existing CALL tools.