Speech as an indicator for psychosocial stress: A network analytic approach DOI Creative Commons
Mitchel Kappen, Kristof Hoorelbeke, Nilesh Madhu

et al.

Behavior Research Methods, Journal Year: 2021, Volume and Issue: 54(2), P. 910 - 921

Published: Aug. 6, 2021

Abstract Recently, the possibilities of detecting psychosocial stress from speech have been discussed. Yet, there are mixed effects and a current lack clarity in relations directions for parameters derived stressed speech. The aim study is – controlled induction experiment to apply network modeling (1) look into unique associations between specific parameters, comparing networks containing fundamental frequency (F0), jitter, mean voiced segment length, Harmonics-to-Noise Ratio (HNR) pre- post-stress induction, (2) examine how changes versus (i.e., change network) each related self-reported negative affect. Results show that similar after before with central role HNR, which shows complex interplay used not impacted by (aim 1). Moreover, we found (consisting pre-post difference values) jitter being positively affect 2). These findings illustrate first time well-controlled but ecologically valid setting different context stress. Longitudinal experimental studies required further investigate these relationships test whether identified paths indicative causal relationships.

Language: Английский

Speech as a promising biosignal in precision psychiatry DOI Creative Commons
Mitchel Kappen, Marie‐Anne Vanderhasselt, George M. Slavich

et al.

Neuroscience & Biobehavioral Reviews, Journal Year: 2023, Volume and Issue: 148, P. 105121 - 105121

Published: March 11, 2023

Health research and health care alike are presently based on infrequent assessments that provide an incomplete picture of clinical functioning. Consequently, opportunities to identify prevent events before they occur missed. New technologies addressing these critical issues by enabling the continual monitoring health-related processes using speech. These a great match for healthcare environment because make high-frequency non-invasive highly scalable. Indeed, existing tools can now extract wide variety health-relevant biosignals from smartphones analyzing person's voice linked biological pathways have shown promise in detecting several disorders, including depression schizophrenia. However, more is needed speech signals matter most, validate against ground-truth outcomes, translate data into biomarkers just-in-time adaptive interventions. We discuss herein describing how assessing everyday psychological stress through help both researchers providers monitor impact has mental physical such as self-harm, suicide, substance abuse, depression, disease recurrence. If done appropriately securely, novel digital biosignal could play key role predicting high-priority outcomes delivering tailored interventions people when need it most.

Language: Английский

Citations

14
Manifestation of depression in speech overlaps with characteristics used to represent and recognize speaker identity DOI Creative Commons

Sri Harsha Dumpala,

Katerina Dikaios,

Sebastián Rodríguez

et al.

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: July 10, 2023

The sound of a person's voice is commonly used to identify the speaker. speech also starting be detect medical conditions, such as depression. It not known whether manifestations depression in overlap with those In this paper, we test hypothesis that representations personal identity speech, speaker embeddings, improve detection and estimation depressive symptoms severity. We further examine changes severity interfere recognition speaker's identity. extract embeddings from models pre-trained on large sample speakers general population without information diagnosis. these for independent datasets consisting clinical interviews (DAIC-WOZ), spontaneous (VocalMind), longitudinal data (VocalMind). use estimates predict presence Speaker combined established acoustic features (OpenSMILE), predicted root mean square error (RMSE) values 6.01 6.28 DAIC-WOZ VocalMind datasets, respectively, lower than alone or alone. When depression, showed higher balanced accuracy (BAc) surpassed previous state-of-the-art performance BAc 66% 64% respectively. Results subset participants repeated samples show identification affected by These results suggest overlaps space. While estimation, deterioration improvement mood may verification.

Language: Английский

Citations

13
Using AI to predict service agent stress from emotion patterns in service interactions DOI
Stefano Bromuri, Alexander P. Henkel, Deniz İren

et al.

Journal of service management, Journal Year: 2020, Volume and Issue: 32(4), P. 581 - 611

Published: Sept. 29, 2020

Purpose A vast body of literature has documented the negative consequences stress on employee performance and well-being. These deleterious effects are particularly pronounced for service agents who need to constantly endure manage customer emotions. The purpose this paper is introduce describe a deep learning model predict in real-time agent from emotion patterns voice-to-voice interactions. Design/methodology/approach was developed identify call center interactions based 363 recorded interactions, subdivided 27,889 manually expert-labeled three-second audio snippets. In second step, deployed period one month be further trained by data collected 40 another 4,672 Findings classifier reached balanced accuracy 68% predicting discrete emotions Integrating binary classification model, it able with 80%. Practical implications Service managers can benefit employing continuously unobtrusively monitor level their numerous practical applications, including early warning systems agents, customized training automatically linking customer-related outcomes. Originality/value present study first document an artificial intelligence (AI)-based that natural (i.e. nonstaged) It pioneer developing smart emotion-based measure agents. Finally, contributes role stress.

Language: Английский

Citations

35
The Feasibility of Wearable and Self-Report Stress Detection Measures in a Semi-Controlled Lab Environment DOI Creative Commons
Sara Aristizabal, Kunjoon Byun, Nadia Wood

et al.

IEEE Access, Journal Year: 2021, Volume and Issue: 9, P. 102053 - 102068

Published: Jan. 1, 2021

Workplace-related stressors, economic strain, and lack of access to educational basic needs have exacerbated feelings stress in the United States. Ongoing can result an increased risk cardiovascular, musculoskeletal, mental health disorders. Similarly, workplace translate a decrease employee productivity higher costs associated with absenteeism organization. Detecting events that correlate during workday is first step addressing its negative effects on wellbeing. Although there are variety techniques for detection using physiological signals, still limited research ability behavioral measures improve performance algorithms. In this study, we evaluated feasibility detecting deep learning, subfield machine small data set consisting electrodermal activity, skin temperature, heart rate measurements, combination self-reported anxiety stress. The model was able detect periods 96% accuracy when combined wearable device survey data, compared dataset alone (88% accuracy). Creating multi-dimensional datasets include both ratings perceived could help stress-inducing at individual level reduce intra-individual variabilities due subjective nature response.

Language: Английский

Citations

32
Detection of mental stress using novel spatio-temporal distribution of brain activations DOI
Debatri Chatterjee, Rahul Gavas, Sanjoy Kumar Saha

et al.

Biomedical Signal Processing and Control, Journal Year: 2023, Volume and Issue: 82, P. 104526 - 104526

Published: Jan. 3, 2023

Language: Английский

Citations

12
Silent Suffering: Using Machine Learning to Measure CEO Depression DOI Creative Commons

Sung-Yuan Cheng,

Nargess M. Golshan

Journal of Accounting Research, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 8, 2025

ABSTRACT We introduce a novel measure of CEO depression by applying machine learning models that analyze vocal acoustic features from CEOs' conference call recordings. Our research was preregistered via the Journal Accounting Research 's registration‐based editorial process. In this study, we validate and examine associated factors. find greater firm risk is positively with depression, whereas higher job demands are negatively depression. Female older CEOs show lower likelihood Using measure, then explore relationship between career outcomes. Although do not any evidence turnover, some turnover‐performance sensitivity among depressed CEOs. also limited compensation pay‐performance for This study provides new insights into mental health

Language: Английский

Citations

0
How Anxiety State Influences Speech Parameters: A Network Analysis Study from a Real Stressed Scenario DOI Creative Commons
Qingyi Wang, Feifei Xu, Xianyang Wang

et al.

Brain Sciences, Journal Year: 2025, Volume and Issue: 15(3), P. 262 - 262

Published: Feb. 28, 2025

Background/Objectives: Voice analysis has shown promise in anxiety assessment, yet traditional approaches examining isolated acoustic features yield inconsistent results. This study aimed to explore the relationship between states and vocal parameters from a network perspective ecologically valid settings. Methods: A cross-sectional was conducted with 316 undergraduate students (191 males, 125 females; mean age 20.3 ± 0.85 years) who completed standardized picture description task while their speech recorded. Participants were categorized into low-anxiety (n = 119) high-anxiety 197) groups based on self-reported ratings. Five parameters—jitter, fundamental frequency (F0), formant frequencies (F1/F2), intensity, rate—were analyzed using analysis. Results: Network revealed robust negative jitter state anxiety, as sole parameter consistently linked total group. Additionally, higher levels associated coupling intensity F1/F2, whereas displayed sparser organization without F1/F2 connection. Conclusions: Anxiety could be recognized by networks ecological The distinct pattern jitter-anxiety connection F1/2 suggest potential markers for assessment. These findings that may directly influence fundamentally restructure relationships among features, highlighting importance of interactions rather than values detection anxiety.

Language: Английский

Citations

0
Robust Relatable Explanations of Machine Learning with Disentangled Cue-specific Saliency DOI

Harshavardhan Sunil Abichandani,

W. Y. Zhang, Brian Y. Lim

et al.

Published: March 19, 2025

Language: Английский

Citations

0
Overview of Voice Biomarkers DOI
Shinichi Tokuno

Published: Jan. 1, 2025

Language: Английский

Citations

0
What Are We Measuring When We Evaluate Digital Interventions for Improving Lifestyle? A Scoping Meta-Review DOI Creative Commons
Rodolfo Castro, Marcelo Ribeiro-Alves, Cátia Cristina Martins de Oliveira

et al.

Frontiers in Public Health, Journal Year: 2022, Volume and Issue: 9

Published: Jan. 3, 2022

Background: Lifestyle Medicine (LM) aims to address six main behavioral domains: diet/nutrition, substance use (SU), physical activity (PA), social relationships, stress management, and sleep. Digital Health Interventions (DHIs) have been used improve these domains. However, there is no consensus on how measure lifestyle its intermediate outcomes aside from measuring each behavior separately. We aimed describe (1) the most frequent domains addressed by DHIs, (2) changes, (3) DHI delivery methods. Methods: followed Preferred Reporting Items for Systematic Reviews Meta-Analyses (PRISMA-ScR) Extension Scoping Reviews. A literature search was conducted using MEDLINE, Cochrane Library, EMBASE, Web of Science publications since 2010. included systematic reviews meta-analyses clinical trials promote health, behavioral, or change. Results: Overall, 954 records were identified, 72 included. Of those, 35 meta-analyses, 58 60 focused PA. Only one review evaluated all simultaneously; 1 five domains; 5 4 14 3 remaining 52 only two The frequently diet/nutrition methods smartphone apps websites. Discussion: concept still unclear fragmented, making it hard evaluate complex interconnections unhealthy behaviors, their impact health. Clarifying this concept, refining operationalization, defining reporting guidelines should be considered as current research priorities. DHIs potential at primary, secondary, tertiary levels prevention-but them are targeting populations. Although important advances made some characteristics, such rate which they become obsolete, will require innovative designs long-term in

Language: Английский

Citations

16