Voice Patterns Research Articles

A generalizable speech emotion recognition model reveals depression and remission.

Affective disorders are associated with atypical voice patterns; however, automated voice analyses suffer from small sample sizes and untested generalizability on external data. We investigated a generalizable approach to aid clinical evaluation of depression and remission from voice using transfer learning: We train machine learning models on easily accessible non-clinical datasets and test them on novel clinical data in a different language. A Mixture of Experts machine learning model was trained to infer happy/sad emotional state using three publicly available emotional speech corpora in German and US English. We examined the model's predictive ability to classify the presence of depression on Danish speaking healthy controls (N = 42), patients with first-episode major depressive disorder (MDD) (N = 40), and the subset of the same patients who entered remission (N = 25) based on recorded clinical interviews. The model was evaluated on raw, de-noised, and speaker-diarized data. The model showed separation between healthy controls and depressed patients at the first visit, obtaining an AUC of 0.71. Further, speech from patients in remission was indistinguishable from that of the control group. Model predictions were stable throughout the interview, suggesting that 20-30 s of speech might be enough to accurately screen a patient. Background noise (but not speaker diarization) heavily impacted predictions. A generalizable speech emotion recognition model can effectively reveal changes in speaker depressive states before and after remission in patients with MDD. Data collection settings and data cleaning are crucial when considering automated voice analysis for clinical purposes.

Asymmetries in Perceptual Adjustments to Non-Canonical Pronunciations

This paper examines two plausible mechanisms supporting sound category adaptation: directional shifts towards the novel pronunciation or a general category relaxation of criteria. Focusing on asymmetries in adaptation to the voicing patterns of English coronal fricatives, we suggest that typology or synchronic experience affect adaptation. A corpus study of coronal fricative substitution patterns confirmed that North American English listeners are more likely to be exposed to devoiced /z/ than voiced /s/. Across two perceptual adaptation experiments, listeners in test conditions heard naturally produced devoiced /z/ or voiced /s/ in critical items within sentences, while control listeners were exposed to identical sentences with canonical pronunciations. Perceptual adaptation was tested via a lexical decision test, with devoiced /z/ or voiced /s/, as well as a novel alveopalatalized pronunciation, to determine whether adaptation was targeted in the direction of the exposed variant or reflected a more general relaxation. Results indicate there was directional and word-specific adaptation for /z/-devoicing with no evidence for generalization. Conversely, there was evidence of /s/-voicing generalizing and eliciting general category relaxation. These results underscore the role of perceptual experiences, and support an evaluation stage in perceptual learning, where listeners assess whether to update a representation.

AN UPSIDE DOWN WORLD? ARTIFICIAL INTELLIGENCE (AI) EXPERIMENTS TO TEACH HUMANS ABOUT EMPATHY: DO ALL ROADS LEAD TO AI?

The fourth industrial revolution is poised to transform the society and the economy we live and work in, which is likely to be well under way by 2030 (Morgan, 2020). Proponents of the fourth industrial revolution postulate that the combination of new technologies, such as machine learning, natural language coding, cloud computing, nanotechnology, sensors, 3D printing, internet of things, and artificial intelligence (AI), holds the transformative power over our societies and economies. Advances made in emotional AI can assist businesses by capturing emotional reactions in real time by decoding facial expressions, measuring neurological immersion levels, analyzing voice patterns, and so on, which would enable organizations to better understand their employees as well as their customers (Purdy et al., 2019).

Gender, Language and Indian Reality Television: Locating Social Stereotypes and Linguistic Sexism

Aim. The reality genre has gained much popularity in last few years across the globe. In India too, this genre has reportedly surpassed other genres in recognition and acceptance. However, its format and content intrigue controversies at both social and academic level, but its mass appeal gets bigger each day. The study examines gender, language and issues of gendered language in Indian reality show Bigg Boss. While keeping in view the format and claim of the genre, the study observes how under thorough surveillance, housemates negotiate with their real self and grapple with their language practice. Influence of neoliberalism on society, culture and identity has been very much discussed, the paper attempts to highlight how this transitions in gender identity is depicted in a show taking non-actors and common people. Concept. For the purpose of the study, scripted transcription of Bigg Boss (Season 11) has been used, wherein all the episodes of have been watched back and forth carefully, and detailed notes prepared for the analysis. Housemates’ language practice, voice pattern and preference of words and statements have been observed. Results and conclusion. The study finds that game reality show Bigg Boss substantially adheres to social stereotypes and standards and while doing so, it also imitates the language practice prevalent everywhere. Although active participation of women in the show is visible butt while exhibiting their true self, sometimes gender prejudice embedded deep inside also comes out, which in case of men is very frequent. The study concludes that as a globally acclaimed genre, reality show, like other genres, is very much commercialized and consumer oriented. Originality. A number of studies related to reality shows have been conducted, but in Indian context, this genre has been inadequately explored. Also, it is very difficult to find studies focusing any specific season of an Indian reality show, so it would not be incorrect to mention that the present study is one of the primary works on this subject, which aims at making some significant contribution related to this genre, in academia.

Pay attention to the speech: COVID-19 diagnosis using machine learning and crowdsourced respiratory and speech recordings

Since the outbreak of COVID-19, many efforts have been made to utilize the respiratory sounds and coughs collected by smartphones for training Machine Learning models to classify and distinguish COVID-19 sounds from healthy ones. Embedding those models into mobile applications or Internet of things devices can make effective COVID-19 pre-screening tools afforded by anyone anywhere. Most of the previous researchers trained their classifiers with respiratory sounds such as breathing or coughs, and they achieved promising results. We claim that using special voice patterns besides other respiratory sounds can achieve better performance. In this study, we used the Coswara dataset where each user has recorded 9 different types of sounds as cough, breathing, and speech labeled with COVID-19 status. A combination of models trained on different sounds can diagnose COVID-19 more accurately than a single model trained on cough or breathing only. Our results show that using simple binary classifiers can achieve an AUC of 96.4% and an accuracy of 96% by averaging the predictions of multiple models trained and evaluated separately on different sound types. Finally, this study aims to draw attention to the importance of the human voice alongside other respiratory sounds for the sound-based COVID-19 diagnosis.

Opportunities and challenges of using biometrics for business: Developing a research agenda

Recently, biometric data generated by fingerprints, hand geometry, heart rate, voice patterns, facial characteristics and expressions, brain activity and body movement has increased in both volume and prominence. Surprisingly, academic business literature has remained relatively silent on the immense potential of biometric data, as well as on the various dangers that come with its collection and usage. This article sets out to (1) detail what biometric data entails and how it may be used, (2) describe opportunities associated with using biometric data in various business applications, (3) discuss challenges related to biometric data collection and usage, privacy and security, storage and safety, and potential for reduced inclusiveness and enhanced biases, and (4) outline related directions for future research.

Analysis of the correlation between the human voice and brain activity

In this paper, we investigated the coupling among the alterations of brain activity and the rhythmic pattern of voice. We benefited from complexity and information concepts and ran the analysis on EEG and voice (audio) signals using sample entropy and Shannon entropy. To change brain activity, we applied four different odors with different complexities on ten subjects (5 M, 5 F). Accordingly, subjects’ voice was changed, and therefore, we evaluated the changes in EEG versus voice signals by calculating their Shannon entropy and sample entropy. The obtained results showed that the variations of complexity (r = 0.8659) and the information content (r = 0.9423) of voice and EEG signals are strongly correlated. This method can be utilized to evaluate the coupling of other biosignals versus brain activity.

Voice Based Disease Identification System

Human voice as well as the sound of the body is used as a clinical method to assess the health condition of an individual. The evaluation of the human voice has risen as a critical field of exploration. Speech analysis fundamentally involves the extraction of certain features from voice signals for generation of voice in alluring pertinence by using reasonable techniques. This paper brings up normal ailments that sway understanding voice patterns in proof for driving research that have affirmed voice modifications as demonstrative manifestations in their respective ailments and also the technique by which voice analysis can be done.

Design of Hand Exoskeleton for Paralysis with Voice Pattern Recognition Control

The paralysis that occurs in the human limbs can be caused by strokes, injuries, age problems, and ligament damage. The purpose of this study is to design a hand exoskeleton as rehabilitation in patients who have had a stroke in hand. The contribution of this research is to design a hand exoskeleton with a control system to control mechanical movements using voice command so that it can be used by patients who have a stroke. To be used by patients who have a stroke, the researcher designed a control system using voice pattern recognition so that patients who have weak myoelectric signals can control the mechanics easily. This device uses the voice recognition module V3 as a voice command to control open and control close mechanical movements. This device is capable of recording and running commands directly by using the push button, which consists of a start, save, reset, open command record, and close command record. In the open command obtained an accuracy value of 97%, the close command obtained an accuracy value of 93%. The results showed that the voice commands given had an average accuracy rate of 95%. The results of this study can be implemented as a rehabilitation device for people who have had a stroke to try to mimic human hand movements.

A machine learning perspective on the emotional content of Parkinsonian speech

Patients with Parkinson's disease (PD) have distinctive voice patterns, often perceived as expressing sad emotion. While this characteristic of Parkinsonian speech has been supported through the perspective of listeners, where both PD and healthy control (HC) subjects repeat the same speaking tasks, it has never been explored through a machine learning modelling approach. Our work provides an objective evaluation of this characteristic of the PD speech, by building a transfer learning system to assess how the PD pathology affects the sadness perception. To do so we introduce a Mixture-of-Experts (MoE) architecture for speech emotion recognition designed to be transferable across datasets. Firstly, by relying on publicly available emotional speech corpora, we train the MoE model and then we use it to quantify perceived sadness in never seen before PD and matched HC speech recordings. To build our models (experts), we extracted spectral features of the voicing parts of speech and we trained a gradient boosting decision trees model in each corpus to predict happiness vs. sadness. MoE predictions are created by weighting each expert's prediction according to the distance between the new sample and the expert-specific training samples. The MoE approach systematically infers more negative emotional characteristics in PD speech than in HC. Crucially, these judgments are related to the disease severity and the severity of speech impairment in the PD patients: the more impairment, the more likely the speech is to be judged as sad. Our findings pave the way towards a better understanding of the characteristics of PD speech and show how publicly available datasets can be used to train models that provide interesting insights on clinical data.

The Active and Passive Voice in the Methods Section of Applied Linguistics Research Articles

The methods section is a critical portion of a research article, which represents the validity and rigor of the overall research design. While previous research has shown a predominance of the passive voice in the research article genre, particularly in the methods section due to its impersonal and explanatory nature (Riley, 1991), recent trends suggests that researchers are increasingly resorting to the active voice in the writing of research article in general (Seoane, 2013). The current study explored whether this trend can be observed in the methods section of applied linguistics research articles and investigated the possible patterns of voice use. The methods sections of 24 articles from international applied linguistics journals were examined to investigate the frequency of voice use and the functional patterns of voice arising, in the corpus. Results suggest that researchers do show alignment with the current trend by actively employing the active voice as well as the passive voice. Functional patterns of voice further revealed that specific discoursal functions within the methods section can be associated with the use of voice. Findings from this study are hoped to serve as the starting point of discussions on the use of voice in particular sections within the research article.

Phonological Variation in Hijazi and Cairene Arabic: The Case of the Uvular Stop /q/

This small scale, qualitatively based, exploratory study aimed at exploring and analysing voice pattern production of eight participants (2 male and 6 female) by eliciting their voice patterns generated from reading forty-three words from a word list and also, from reading a short paragraph. Furthermore, the study aimed at exploring the phonological variation in the spoken Arabic of two regions: Jeddah, Saudi Arabia (Hijazi Arabic—HA) and Cairo, Egypt (Cairene Arabic—CA) and also, examines the role of age, sex, educational level and region on this variation. In particular, the study explored the correlation of voice patterns production as they relate to the non-linguistic factors such as age, gender and educational level. The analysis of the gathered qualitative data was carried out by transcribing the recorded data and coding each voice pattern in question. The results indicated that there is more variation in CA speakers than in HA speakers and also, there are some gender influences on phonological production.

Development of SelfSpeaking Body Weight Scale for Visually Impaired People in Tanzania

Speaking weight scale is an important low vision health aid that measures and announces the measured weight. It is valuable in numerous applications such as Bathroom scale, Kitchen scale, and more. Different talking scales have been developed for the blind community. Many talking scales have language options for English, German, French, or Spanish. However, only limited work exists for Self speaking visually impaired community in EAC given the fact that no talking scale can announce weight in Self which is the common language in EAC. Therefore, this project aims to develop a Self-speaking weighing machine to assist visually impaired people in Tanzania. The developed device is divided into two major parts. On the front end of the design, sensors are used to capture weight parameters. The captured values are mapped onto a sequence of voice patterns. The back-end consists of transferring a sequence of voice patterns to a loudspeaker whereby the voice patterns are stored on an SD card. Finally, the developed device has been evaluated on several objects with known weights. The results show that the developed device accurately measures weight, displays weight, and announces it in the Self language. However, blind people still need assistance from sighted persons to be directed to the scale’s platform. The developed device has great potential as a low vision health aid for Self speakers. Moreover, the features of this device can be further improved by integrating iBeacon technology to increase the autonomy of blind people to use the scale and navigate to the device’s location safely.

Construções reflexiva, recíproca e média de clítico nulo no português brasileiro reconceptualização de eventos e emergência de uma nova construção

tuguese se constructions, posited in the transitive continuum, have a constructional counterpart in which the clitic is absent. The null clitic construction, observed in all the seconstructions (i.e. reflexive, reciprocal, middle, anticausative, passive and impersonal) is more frequently used in Brazilian Portuguese (BP) than in European Portuguese (EP). The phenomenon has largely been studied from a morphosyntatic lens or as a result of an ongoing deletion of clitics in BP, shying away from the possible implications in terms of the semantic differentiation between overt and null se constructions. This chapter focuses on reflexive, reciprocal and middle se constructions and aims to investigate what factors determine the choice between overt seconstructions and their null counterpart. Based on an extensive usage-feature and profile-based analysis, and using multivariate statistical methods, we show that reflexive, reciprocal and middle null se constructions are associated with a reconceptualization of an event as non-energetic or absolute, profiling the result of the event. On the other hand, the overt counterpart profiles the moment of change, construing the event as energetic. Reflexive and reciprocal constructions are more frequently encoded by an overt se construction whereas middle construction (in all its subcategories) is more frequently encoded by the null se construction. The study concludes that null reflexive, reciprocal and middle se constructions are new constructions semantically differentiated from overt se constructions, which, we argue, has wider implications, namely for reconceptualization of voice patterns in BP which tend towards ergativization.

Optimization Research on Abnormal Diagnosis of Transformer Voiceprint Recognition based on Improved Wasserstein GAN

Transformer is an important infrastructure equipment of power system, and fault monitoring is of great significance to its operation and maintenance, which has received wide attention and much research. However, the existing methods at home and abroad are based on model analysis for detection and diagnosis techniques, and their application is limited by the maturity and applicability of expert knowledge. In recent years, deep learning artificial intelligence has made important breakthroughs, however, the relevant research is still mainly focused on the field of speech, and intelligent diagnostic techniques on machine voice patterns have just begun. Therefore, this article adopts an unsupervised learning anomaly detection method based on WGAN-GP to diagnose the electrical and mechanical anomalies of transformer equipment. The experimental results show that the method can effectively identify transformer anomalies and provide an idea for artificial intelligence-based machine voice pattern recognition.

Designing a Voice-controlled Wheelchair for Persian-speaking Users Using Deep Learning Networks with a Small Dataset

With the advancement of technology, the demand for improving the quality of life of the elderly and disabled has increased and their hope to overcome their problem is realized by using advanced technologies in the field of rehabilitation. Many existing electrical and electronic devices can be turned into more controllable and more functional devices using artificial intelligence. In every society, some spinal disabled people lack physical and motor abilities such as moving their limbs and they cannot use the normal wheelchair and need a wheelchair with voice control. The main challenge of this project is to identify the voice patterns of disabled people. Audio classification is one of the challenges in the field of pattern recognition. In this paper, a method of classifying ambient sounds based on the sound spectrogram, using deep neural networks is presented to classify Persian speakers sound for building a voice-controlled intelligent wheelchair. To do this, we used Inception-V3 as a convolutional neural network which is pretrained by the ImageNet dataset. In the next step, we trained the network with images that are generated using spectrogram images of the ambient sound of about 50 Persian speakers. The experimental results achieved a mean accuracy of 83.33%. In this plan, there is the ability to control the wheelchair by a third party (such as spouse, children or parents) by installing an application on their mobile phones. This wheelchair will be able to execute five commands such as stop, left, right, front and back.

Penerapan Hidden Markov Model (HMM) dan Mel-Frequency Cesptral Coefficients (MFCC) pada E-Learning Bahasa Madura untuk Anak Usia Dini

Bahasa Madura is a regional language used in Madura island. This language has many variations of pronunciation and dialect that makes it not easy to learn, even by the local people especially children. There hasn’t been any interesting learning media to learn Bahasa Madura so far. In fact, a fun learning activity is needed to help children to enhance their ability in pronouncing animals’ names, numbers, fruits and things in Bahasa Madura. Thus, it’s considered important to create Bahasa Madura e-learning by implementing the recognition of voice patterns in order to make it easier for the children to learn Bahasa Madura which has several variations of pronunciation only for one single object. This Bahasa Madura e-learning application for young learners is used to introduce Bahasa Madura vocabularies by recognizing the voice pattern recordings which have been processed through MFCC technique as the extracted voice features and HMM as the learning techniques. The implementation of MFCC and HMM as the learning tool to introduce the pronunciation of regional language vocabularies especially Bahasa Madura has never been done before. Therefore, this research is expected to help the young learners to be able to pronounce Bahasa Madura vocabularies properly. In this study, a number of young learners’ voices were recorded and were set as the trial data. Only the proper voice data that were used—voice data that were considered to be pronounced correctly. The trial method was done through one-single model and multi-model. After doing several simultaneous trials, the result showed the accuracy level. The average accuracy level for one-single model system was 73% (with the highest accuracy reached 75%) and the average accuracy level for multi-model system was 80% (with the highest accuracy reached 81%).

A “Proto-Theme” in Some of J. S. Bach’s Fugal Works

This article focuses on a pattern of harmony and voice leading that appears, elaborated, in several fugue expositions (and generically related passages) by J. S. Bach. I contextualize this pattern in examples of “schematic” Baroque fugal writing and suggest that for Bach it functions as a sort of “proto-theme,” with respect to which certain expositions stand as “variations.” These expositions eschew normative I–V–I progressions ofStufenand thus offer a challenge to William Renwick’s exhaustive methodology for fugue analysis. Bach’s variations may respond motivically and energetically to a sense of retrogression implicit the proto-theme, as I demonstrate with an analysis of the Fugue in A minor from Book II ofThe Well-Tempered Clavier, BWV 889.

Association Between Electroencephalogram-Derived Sleep Measures and the Change of Emotional Status Analyzed Using Voice Patterns: Observational Pilot Study.

BackgroundMeasuring emotional status objectively is challenging, but voice pattern analysis has been reported to be useful in the study of emotion.ObjectiveThe purpose of this pilot study was to investigate the association between specific sleep measures and the change of emotional status based on voice patterns measured before and after nighttime sleep.MethodsA total of 20 volunteers were recruited. Their objective sleep measures were obtained using a portable single-channel electroencephalogram system, and their emotional status was assessed using MIMOSYS, a smartphone app analyzing voice patterns. The study analyzed 73 sleep episodes from 18 participants for the association between the change of emotional status following nighttime sleep (Δvitality) and specific sleep measures.ResultsA significant association was identified between total sleep time and Δvitality (regression coefficient: 0.036, P=.008). A significant inverse association was also found between sleep onset latency and Δvitality (regression coefficient: –0.026, P=.001). There was no significant association between Δvitality and sleep efficiency or number of awakenings.ConclusionsTotal sleep time and sleep onset latency are significantly associated with Δvitality, which indicates a change of emotional status following nighttime sleep. This is the first study to report the association between the emotional status assessed using voice pattern and specific sleep measures.

M248. VOICE PATTERNS IN SCHIZOPHRENIA: A CROSS-LINGUISTIC REPLICATION OF PREVIOUS META-ANALYTIC FINDINGS

BackgroundSchizophrenia (SCZ) has been associated to distinctive voice since its first definitions. Distinctive voice patterns are often associated with core negative symptoms and with social impairment. They may thus represent markers of the disorder. A recent meta-analysis identified weak atypicalities for pitch variability, and stronger atypicalities in duration (speech percentage, pause duration and speech rate). However, heterogeneity across studies was large, most of the studies underpowered (small sample and no repeated measures) and replications across studies almost nonexistent. In addition, there is a lack of cross-linguistic studies comparing voice and linguistic patterns in SCZ across different languages to assess whether the patterns are distinctive of SCZ in general, or specific to linguistic and/or cultural groups. In the present study, we aim to advance the understanding of voice patterns in SCZ by collecting and analyzing a cross-linguistic corpus of repeated voice measures. Such corpus enables us to systematically assess the replicability of previous meta-analytic results, better accounting for between and within participant variability, as well as cross-linguistic differences.MethodsWe collected a Danish (DK), Chinese (CH) and Japanese (JP) cross-linguistic dataset involving 163 participants with SCZ (105 DK, 51 CH, 7 JP) and 173 matched controls (HC) (117 DK, 43 CH, 13 JP) for a total of 3851 audio-recordings. Data were collected using the Animated Triangle 1 2020 Congress of the Schizophrenia International Research Society Task. Voice recordings were preprocessed using consolidated algorithms (Covarep, Praat) to extract the following features, in order to compare results with the effect sizes (ES) of previous meta-analysis (MA): 1) Duration measures (speech rate, duration of utterance, number of pauses, pause duration), as well as 2) pitch and intensity (mean and variability). To investigate differences between SCZ and HC, we ran multilevel regression models with the acoustic feature as outcome, diagnosis (SZ, HC) and language (DK, JP, CH) as predictors, and varying effects by participant and corpus. Predictors were scaled in order to allow comparison with meta-analysis ES.ResultsWe were only able to partially replicate previous findings. The meta-analysis found: 1) lower pitch variability, replicated for JP only (β= -1.25, SE = 0.37, p < .001); 2) lower speech rate replicated for DK only (β= -0.23, SE = .08, p < .01); 3) increased pause duration replicated for DK (β= 0.29, SE = .08, p < .001) and JP (β= 0.59, SE = .30, p < .05); 4) lack of evidence for atypical number of pauses replicated for DK, JP and CH; 5) lack of evidence for atypical duration of utterance replicated for CH and JP (DK presented higher duration: β= 0.01, SE = 0.01, p < .01); 6) lower proportion of spoken time, not replicated; 7) lack of evidence for pitch mean, replicated for DK, but higher in CH (β= 0.37, SE = .18, p < .05), and lower in JP (β= -1.46, SE = .41, p < .001).DiscussionWe found only partial replication of previous meta-analytic findings for reduced pitch variability, increased pause duration and lower speech rate, with ES generally smaller than in previous meta-analysis. On the contrary, we were not able to replicate previous findings of lower proportion of spoken time. Estimations of ES were largely affected by different languages, and replications held only for specific languages (pitch variability for JP, speech rate for DK, and pause duration for DK and JP). This indicates the important role that linguistic factors may play in originating vocal patterns in SCZ. Voice patterns seem not to be distinctive of SCZ in general, but bounded to linguistic/cultural differences. Future studies should better investigate how different acoustic and linguistic features interact in originating atypical voice patterns in SCZ.