Search

This paper focusses on using neural network models to predict the age of social media users based on their voice recordings. The objective is to identify potential risky interactions between minors and adults by comparing the declared and predicted age groups of the users. The paper addresses the selection and training of suitable models and evaluates their effectiveness in age prediction. The results are demonstrated in sample data, where performance metrics are analysed, and possible limitations of the method are identified. Finally, the implications of the results for the safety of minors on social networks are discussed, and suggestions for future research in this area are provided.

In recent years, the growing interest in automatic speech recognition (ASR) has been driven by its wide-ranging applications across various domains. Integrating speech recognition technologies into smart systems highlights the pivotal role of human-machine interaction. This study introduces a robust ASR system that leverages convolutional neural networks (CNNs) in conjunction with Mel-frequency cepstral coefficients (MFCCs). The model's architecture was improved by extensively examining hyperparameters, effectively recognizing ten different spoken commands. The model conducted training and evaluation using the Google Speech dataset, comprising 65,000 audio clips collected from a wide range of speakers across the globe. This dataset accurately represents the natural variations in speech found in real-world scenarios. The design comprises eight storage layers, encompassing convolutional and fully connected layers. It consists of a total of 183,345 weights and utilizes ReLU activation. It is worth mentioning that the average F1-score obtained during the training, validation, and testing stages is 99.06 %, 94.68%, and 95.27%, respectively. Furthermore, the proposed model exhibits about 1.3% improvement in experimental test accuracy over existing methods, confirming its effectiveness in real-world applications.

Voice pathologies significantly affect social interactions and quality of life. Traditional diagnostic techniques, such as a laryngoscopy, are invasive and cumbersome. This paper proposes a noninvasive deep learning approach that utilizes voice signal analysis for diagnosis and reduces discomfort and examination time. The main contributions of this study include the implementation of few-shot learning (FSL) in voice pathology detection, addressing data scarcities and class imbalances through its integration with transfer learning. The study employs a metric-based FSL approach to enhance model generalization with limited data and validates this method using both voice and electroglottographic (EGG) signals, thereby improving the diagnostic accuracy and clinical applicability. The effectiveness of the proposed method in voice pathology detection has been demonstrated through experiments using voice and EGG signals, outperforming current FSL algorithms. Using a metric-based FSL framework, test accuracies of 73.7% and 82.6% for voice and EGG data, respectively, have been achieved, highlighting the robustness of the method, particularly for more informative EGG signals. Future work includes further development and validation in real clinical settings.

Emotional communication is a multi-modal phenomenon involving posture, gestures, facial expressions, and the human voice. Affective states systematically modulate the acoustic signals produced during speech production through the laryngeal muscles via the central nervous system, transforming the acoustic signal into a means of affective transmission. Additionally, a substantial body of research in sonobiology has shown that audible acoustic waves (AAW) can affect cellular dynamics. This pilot study explores whether the physical–acoustic changes induced by gratitude states in human speech could influence cell proliferation and Ki67 expression in non-auditory cells (661W cell line). We conduct a series of assays, including affective electroencephalogram (EEG) measurements, an affective text quantification algorithm, and a passive vibro-acoustic treatment (PVT), to control the CO2 incubator environment acoustically, and a proliferation assay with immunolabeling to quantify cell dynamics. Although a larger sample size is needed, the hypothesis that emotions can act as biophysical agents remains a plausible possibility, and feasible physical and biological pathways are discussed. In summary, studying the impact of gratitude AAW on cell biology represents an unexplored research area with the potential to enhance our understanding of the interaction between human cognition and biology through physics principles.

The problem of varying length recordings is a well-known issue in paralinguistics. We investigated how to resolve this problem using the bag-of-audio-words feature extraction approach. The steps of this technique involve preprocessing, clustering, quantization and normalization. The bag-of-audio-words technique is competitive in the area of speech emotion recognition, but the method has several parameters that need to be precisely tuned for good efficiency. The main aim of our study was to analyse the effectiveness of bag-of-audio-words method and try to find the best parameter values for emotion recognition. We optimized the parameters one-by-one, but built on the results of each other. We performed the feature extraction, using openSMILE. Next we transformed our features into same-sized vectors with openXBOW, and finally trained and evaluated SVM models with 10-fold-crossvalidation and UAR. In our experiments, we worked with a Hungarian emotion database. According to our results, the emotion classification performance improves with the bag-of-audio-words feature representation. Not all the BoAW parameters have the optimal settings but later we can make clear recommendations on how to set bag-of-audio-words parameters for emotion detection tasks.

Acoustic simulation of sound propagation inside the vocal tract is a key element of speech research, especially for articulatory synthesis, which allows one to relate the physics of speech production to other fields of speech science, such as speech perception. Usual methods, such as the transmission line method, have a very low computational cost and perform relatively good up to 4–5 kHz, but are not satisfying above. Fully numerical 3D methods such as finite elements achieve the best accuracy, but have a very high computational cost. Better performances are achieved with the state of the art semi-analytical methods, but they cannot describe the vocal tract geometry as accurately as fully numerical methods (e.g. no possibility to take into account the curvature). This work proposes a new semi-analytical method that achieves a better description of the three-dimensional vocal-tract geometry while keeping the computational cost substantially lower than the fully numerical methods. It is a multimodal method which relies on two-dimensional finite elements to compute transverse modes and takes into account the curvature and the variations of cross-sectional area. The comparison with finite element simulations shows that the same degree of accuracy (about 1% of difference in the resonance frequencies) is achieved with a computational cost about 10 times lower.

Il rango assunto dalla musica strumentale nell’estetica del romanticismo tedesco si accompagna, come è noto, nella prassi musicale dell’epoca a un perfezionamento della tecnica di costruzione degli strumenti musicali. Anche in letteratura l’interesse per la meccanica strumentale si riscontra in diversi autori, collegandosi alla fascinazione/demonizzazione nei confronti della tecnologia in generale. Sempre più spesso agli strumenti viene attribuita una voce che si pone in concorrenza con la voce umana, ora come imitazione, ora come superamento e potenziamento. In particolare nell’opera di E. T. A. Hoffmann, lo scrittore che più di ogni altro dispone di competenze musicali professionali, strumento e voce umana (quest’ultima intesa come manifestazione immediata dell’anima) sono oggetto di una serie di scambi, mutazioni e sostituzioni che creano un fitto intreccio fra ciò che si presenta come ‘naturale’ e ciò che è artificio (o illusionismo). Interessanti in particolare i casi del violino nel racconto Rat Krespel e dell’Unsichtbares Mädchen nel romanzo Kater Murr.

Rapid population growth and the urbanization of modern environments are markedly increasing human-wildlife conflict. Wild animals in urban landscapes can benefit from exploiting human resources, but are also exposed to increased risk of human-caused injury, which should favor the ability to perceive and respond to human cues. Although it is well known that domesticated animals use human cues that may indicate threats, less is known about wild animals living in urban environments. Herring gulls (Larus argentatus) in urban landscapes have adapted kleptoparasitic behaviors to obtain human food, often resulting in negative interactions with humans. Here we quantified both the behavioral and physiological responses of free-living urban herring gulls to human shouting. We presented urban gulls with a fake human food item and played back recordings of either a man shouting, a natural stressor (i.e., conspecific alarm call), or a neutral stimulus (i.e., robin song). We recorded behavioral responses and used non-invasive infrared thermography to measure eye-region surface temperature changes associated with the avian physiological stress response. We found that gulls exposed to shouting and to conspecific alarm calls showed similar changes in behavior (indicating high levels of vigilance) and eye-region surface temperature (indicating physiological stress). Both responses were significantly stronger than the responses to robin song. Additionally, the behavioral and physiological responses were positively correlated across individuals. Our results demonstrate that urban-dwelling gulls respond to human shouting and conspecific alarm calls in a similar way, and suggest that infrared thermography is a viable technique to monitor stress responses in free-living birds.

The field of psychology of singing is the least examined area so far, so in the research study we present the research survey from the field of human voice and singing. Music allows the access to the experience of emotions, which can then manifest itself in various changes at a psychosomatic level. New research findings can therefore be used very effectively in the field of music therapy or singing. The anthropological idea of this phenomenon is based on the fact that the human voice is an extremely sensitive physical and emotional tool through which it is possible to act on the body, psyche and spirit, in terms of complex personality development and to discover human contact with its deeper, archetypal components. The sound experience of singing can touch a person’s most basic existential experiences. In this study we want to reveal the underappreciated possibilities of using the human voice and singing in the field of music therapy and point out the purposeful development of vocal dispositions, in concord with the core of human personality (conscious and unconscious). Mainly owing to strong emotions, it has a significant impact on harmonisation of hemispheres, stimulation of all mental functions that leads to spontaneous revitalisation of the psyche and body and it subsequently leads to improved quality of life.

IntroductionTo self-monitor asthma symptoms, existing methods (e.g. peak flow metre, smart spirometer) require special equipment and are not always used by the patients. Voice recording has the potential to generate surrogate measures of lung function and this study aims to apply machine learning approaches to predict lung function and severity of abnormal lung function from recorded voice for asthma patients.MethodsA threshold-based mechanism was designed to separate speech and breathing from 323 recordings. Features extracted from these were combined with biological factors to predict lung function. Three predictive models were developed using Random Forest (RF), Support Vector Machine (SVM), and linear regression algorithms: (a) regression models to predict lung function, (b) multi-class classification models to predict severity of lung function abnormality, and (c) binary classification models to predict lung function abnormality. Training and test samples were separated (70%:30%, using balanced portioning), features were normalised, 10-fold cross-validation was used and model performances were evaluated on the test samples.ResultsThe RF-based regression model performed better with the lowest root mean square error of 10·86. To predict severity of lung function impairment, the SVM-based model performed best in multi-class classification (accuracy = 73.20%), whereas the RF-based model performed best in binary classification models for predicting abnormal lung function (accuracy = 85%).ConclusionOur machine learning approaches can predict lung function, from recorded voice files, better than published approaches. This technique could be used to develop future telehealth solutions including smartphone-based applications which have potential to aid decision making and self-monitoring in asthma.

The prominent and early presence of dysphonia is considered a valuable marker for differentiation of idiopathic Parkinson's disease and parkinsonian syndromes. Objective quantification of vibrational regimes represented by modal and subharmonic vibrations may thus be vital for improving accuracy of diagnostic decision. The rationale for analyzing vibrational regimes is that abnormal subharmonic vibrations might be the key factor causing dysphonia in parkinsonian syndromes. This study introduces a new fully automated methodology based on robust pitch tracker for decoupling vibrations controlled by laryngeal muscles from the effect of subharmonics that provides distinguishing features of Parkinson's disease and atypical parkinsonian syndromes. We tested the method on resynthesized signals with known parameters and demonstrated that vibrations controlled by laryngeal muscles as well as subharmonics can be detected reliably with a precision that outperforms available technologies. We analysed 337 sustained vowels of 22 patients with PD, 21 patients with multiple system atrophy, 18 patients with progressive supranuclear palsy and 22 healthy controls. Our results showed that subharmonics are more prominent in atypical parkinsonian syndromes compared to Parkinson's disease. Also, increased modulation by laryngeal muscles appears to be a distinctive symptom of multiple system atrophy. Developed algorithm and proposed resynthesized voice signals provide further critical step to understanding and evaluation of dysphonia in Parkinsonism.

Patients with Parkinson's disease (PD) may have difficulties in speaking because of reduced coordination of the muscles that control breathing, phonation, articulation, and prosody. Symptoms that may occur are weakening of the volume of the voice, voice monotony, changes in the quality of the voice, the speed of speech, uncontrolled repetition of words, and difficult speech intelligibility. To date, evaluation of the speech intelligibility is performed based on the unified PD rating scale. Specifically, section 3.1 (eloquence) of the cited scale provides the specialist with some tips to evaluate the patient's speech ability. With the aim of evaluating the speech intelligibility by measuring the variation in parameters in an objective manner, we show that a speech-to-text (STT) system could help specialists to obtain an accurate and objective measure of speech, phrase, and word intelligibility in PD. STT systems are based on methodologies and technologies that enable the recognition and translation of spoken language into text by computers and computerized devices. We decided to base our study on Google STT conversion. We expand Voxtester, a software system for digital assessment of voice and speech changes in PD, in order to perform this study. No previous studies have been presented to address the mentioned challenges based on STT. The experiments here presented are related with detection/classification between pathological speech from patients with PD and regular speech from healthy control group. The results are very interesting and are an important step toward assessing the intelligibility of the speech of PD patients.

Perceived vocal attractiveness and measured sex-dimorphic vocal parameters are both associated with underlying individual qualities. Research tends to focus on speech but singing is another highly evolved communication system that has distinct and universal features with analogs in other species, and it is relevant in mating. Both speaking and singing voice provides relevant information about its producer. We tested whether speech and singing function as “backup signals” that indicate similar underlying qualities. Using a sample of 81 men and 86 women from Brazil and the Czech Republic, we investigated vocal attractiveness rated from speech and singing and its association with fundamental frequency (F0), apparent vocal tract length (VTL), body characteristics, and sociosexuality. F0, VTL, and rated attractiveness of singing and speaking voice strongly correlated within the same individual. Lower-pitched speech in men, higher-pitched speech and singing in women, individuals who like to sing more, and singing of individuals with a higher pitch modulation were perceived as more attractive. In men, physical size positively predicted speech and singing attractiveness. Male speech but not singing attractiveness was associated with higher sociosexuality. Lower-pitched male speech was related to higher sociosexuality, while lower-pitched male singing was linked to lower sociosexuality. Similarly, shorter speech VTL and longer singing VTL predicted higher sociosexuality in women. Different vocal displays function as “backup signals” cueing to attractiveness and body size, but their relation to sexual strategies in men and women differs. Both singing and speech may indicate evolutionarily relevant individual qualities shaped by sexual selection.

The flow-induced sound sources of human voice production are investigated based on a validated voice model. This analysis is performed using a hybrid aeroacoustic workflow based on the perturbed convective wave equation. In the first step, the validated 3D incompressible turbulent flow simulation is computed by the finite volume method using STARCCM+. In a second step, the aeroacoustic sources are evaluated and studied in detail. The formulation of the sound sources is compared to the simplification (neglecting the convective sources) systematically using time-domain and Fourier-space analysis. Additionally, the wave equation is solved with the finite element solver openCFS to obtain the 3D sound field in the acoustic far-field. During the detailed effect analysis, the far-field sound spectra are compared quantitatively, and the flow-induced sound sources are visualized within the larynx. In this contribution, it is shown that the convective part of the sources dominates locally near the vocal folds (VFs) while the local time derivative of the incompressible pressure is distributed in the whole supra-glottal area. Although the maximum amplitude of the time derivative is lower, the integral contribution dominates the sound spectrum. As a by-product of the detailed perturbed convective wave equation source study, we show that the convective source term can be neglected since it only reduces the validation error by 0.6%. Neglecting the convective part reduces the algorithmic complexity of the aeroacoustic source computation of the perturbed convective wave equation and the stored flow data. From the source visualization, we learned how the VF motion transforms into specific characteristics of the aeroacoustic sources. We found that if the VFs are fully closing, the aeroacoustic source terms yield the highest dynamical range. If the VFs are not fully closing, VFs motion does not provide as much source energy to the flow-induced sound sources as in the case of a healthy voice. As a consequence of not fully closing VFs, the cyclic pulsating velocity jet is not cut off entirely and therefore turbulent structures are permanently present inside the supraglottal region. These turbulent structures increase the broadband component of the voice signal, which supports research results of previous studies regarding glottis closure and insufficient voice production.

This paper investigates the use of a simplified set of Arabic phonemes in an Arabic Speech Recognition system applied to Holy Quran. The CMU Sphinx 4 was used to train and evaluate a language model for the Hafs narration of the Holy Quran. The building of the language model was done using a simplified list of Arabic phonemes instead of the mainly used Romanized set in order to simplify the process of generating the language model. The experiments resulted in very low Word Error Rate (WER) reaching 1.5% while using a very small set of audio files during the training phase when using all the audio data for both the training and the testing phases. However, when using 90% and 80% of the training data, the WER obtained was respectively 50.0% and 55.7%.

The intensive research of speech emotion recognition introduced a huge collection of speech emotion features. Large feature sets complicate the speech emotion recognition task. Among various feature selection and transformation techniques for one-stage classification, multiple classifier systems were proposed. The main idea of multiple classifiers is to arrange the emotion classification process in stages. Besides parallel and serial cases, the hierarchical arrangement of multi-stage classification is most widely used for speech emotion recognition. In this paper, we present a sequential-forward-feature-selection-based multi-stage classification scheme. The Sequential Forward Selection (SFS) and Sequential Floating Forward Selection (SFFS) techniques were employed for every stage of the multi-stage classification scheme. Experimental testing of the proposed scheme was performed using the German and Lithuanian emotional speech datasets. Sequential-feature-selection-based multi-stage classification outperformed the single-stage scheme by 12–42 % for different emotion sets. The multi-stage scheme has shown higher robustness to the growth of emotion set. The decrease in recognition rate with the increase in emotion set for multi-stage scheme was lower by 10–20 % in comparison with the single-stage case. Differences in SFS and SFFS employment for feature selection were negligible.

Acoustic analysis is used to assist differential diagnosis, documentation and evaluation of treatment for voice disorders. Clinical data has shown that Jitter, Shimmer, Mean Pitch and Harmonic Noise Ratio are the indices of voice pathology. A voice with some periodicity can now be analysed with a computerised acoustic analyser, a relatively newer technique that can be widely used in clinical practice. Objectives: To create a database of normal voices, analyse and identify different of these voices and hence identify benchmarks of normal voices. Materials and Methods: Voice samples of 458 normal males and 542 normal females aged between 18 to 28 years were collected using a sustained vowel /a/ which was recorded and analysed using freely downloadable software “PRAAT”. The parameters like Jitter, Shimmer, Harmonic to Noise Ratio and Pitch were derived and mean, SD and range of voice parameters were calculated. Results: In males, the value of parameters were mean pitch (137.05), jitter (0.011), shimmer (0.08) and Harmonic to Noise Ratio (20.48). In females, the parameters were mean pitch (234.27), jitter (0.01), shimmer (0.08) and harmonics to noise ratio(21.73). Conclusion: Voices can be objectively analysed using acoustic parameters like mean pitch, jitter, shimmer and harmonic to noise ratio. A large database yields more reliable normative parameters. Institutions should develop their own standard protocol for selection of subjects, recording of voices and their analysis

El simple hecho de enseñar es todo un arte y como todo arte demanda una búsqueda constante que implica un esfuerzo por tratar de mejorar esta práctica tan cotidianamente necesaria, que al fin y al cabo es la que nos ha formado y nos sigue estimulando. Lo que trataré a continuación es una descripción clara y con detalles de importancia sobre La Voz humana para la acción pedagógica, considerando especialmente el nuevo momento generacional en que estamos ya inmersos en este siglo lleno de alta tecnología y con miras a una mejoría educativa a través de una excelencia escénica y de dicción de los educadores ante su alumnado.