29 September 2011 in IEEE Transactions on Information Technology in Biomedicine
Chronic stress is endemic to modern society. However, as it is unfeasible for physicians to continuously monitor stress levels, its diagnosis is nontrivial. Wireless body sensor networks offer opportunities to ubiquitously detect and monitor mental stress levels, enabling improved diagnosis, and early treatment. This article describes the development of a wearable sensor platform to monitor a number of physiological correlates of mental stress. We discuss tradeoffs in both system design and sensor selection to balance information content and wearability. Using experimental signals collected from the wearable sensor, we describe a selected number of physiological features that show good correlation with mental stress. In particular, we propose a new spectral feature that estimates the balance of the autonomic nervous system by combining information from the power spectral density of respiration and heart rate variability. We validate the effectiveness of our approach on a binary discrimination problem when subjects are placed under two psychophysiological conditions: mental stress and relaxation. When used in a logistic regression model, our feature set is able to discriminate between these two mental states with a success rate of 81% across subjects.
5 May 2025 in IEEE Transactions on Audio, Speech and Language Processing
Transformer and its derivatives have achieved success in diverse tasks across computer vision, natural language processing, and speech processing. To reduce the complexity of computations within the multi-head self-attention mechanism in Transformer, Selective State Space Models (i.e., Mamba) were proposed as an alternative. Mamba exhibited its effectiveness in natural language processing and computer vision tasks, but its superiority has rarely been investigated in speech signal processing. This paper explores solutions for applying Mamba to speech processing by discussing two typical speech processing tasks: speech recognition, which requires semantic and sequential information, and speech enhancement, which focuses primarily on sequential patterns. The experimental results confirm that bidirectional Mamba (BiMamba) consistently outperforms vanilla Mamba, highlighting the advantages of a bidirectional design for speech processing. Moreover, experiments demonstrate the effectiveness of BiMamba as an alternative to the self-attention module in the Transformer model and its derivates, particularly for the semantic-aware task. The crucial technologies for transferring Mamba to speech are then summarized in ablation studies and the discussion section, offering insights for extending this research to a broader scope of tasks.
9 January 2017 in IEEE Journal of Biomedical and Health Informatics
FMFathima Lamana Mulaffer
Effective sleep analysis is hampered by the lack of automated tools catering to disordered sleep patterns and cumbersome monitoring hardware. In this paper, we apply deep learning on a set of 57 EEC features extracted from a maximum of two EEC channels to accurately differentiate between patients with insomnia or controls with no sleep complaints. We investigated two different approaches to achieve this. The first approach used EEC data from the whole sleep recording irrespective of the sleep stage (stage-independent classification), while the second used only EEC data from insomnia-impacted specific sleep stages (stage-dependent classification). We trained and tested our system using both healthy and disordered sleep collected from 41 controls and 42 primary insomnia patients. When compared with manual assessments, an NREM + REM based classifier had an overall discrimination accuracy of 92% and 86% between two groups using both two and one EEC channels, respectively. These results demonstrate that deep learning can be used to assist in the diagnosis of sleep disorders such as insomnia.
9 October 2018 in International Journal of Speech-Language Pathology
Purpose: To assist in remote treatment, speech-language pathologists (SLPs) rely on mobile games, which though entertaining, lack feedback mechanisms. Games integrated with automatic speech recognition (ASR) offer a solution where speech productions control gameplay. We therefore performed a feasibility study to assess children’s and SLPs’ experiences towards speech-controlled games, game feature preferences and ASR accuracy. Method: Ten children with childhood apraxia of speech (CAS), six typically developing (TD) children and seven SLPs trialled five games and answered questionnaires. Researchers also compared the results of ASR to perceptual judgment. Result: Children and SLPs found speech-controlled games interesting and fun, despite ASR–human disagreements. They preferred games with rewards, challenge and multiple difficulty levels. Automatic speech recognition–human agreement was higher for SLPs than children, similar between TD and CAS and unaffected by CAS severity (77% TD, 75% CAS – incorrect; 51% TD, 47% CAS, 71% SLP – correct). Manual stop recording yielded higher agreement than automatic. Word length did not influence agreement. Conclusion: Children’s and SLPs’ positive responses towards speech-controlled games suggest that they can engage children in higher intensity practice. Our findings can guide future improvements to the ASR, recording methods and game features to improve the user experience and therapy adherence.
11 July 2018 in International Journal of Speech-Language Pathology
Purpose: A systematic search and review of published studies was conducted on the use of automated speech analysis (ASA) tools for analysing and modifying speech of typically-developing children learning a foreign language and children with speech sound disorders to determine (i) types, attributes, and purposes of ASA tools being used; (ii) accuracy against human judgment; and (iii) performance as therapeutic tools. Method: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were applied. Across nine databases, 32 articles published between January 2007 and December 2016 met inclusion criteria: (i) focussed on children’s speech; (ii) tools used for speech analysis or modification; and (iii) reporting quantitative data on accuracy. Result: Eighteen ASA tools were identified. These met the clinical threshold of 80% agreement with human judgment when used as predictors of intelligibility, impairment severity, or error category. Tool accuracy was typically <80% accuracy for words containing mispronunciations. ASA tools have been used effectively to improve to children’s foreign language pronunciation. Conclusion: ASA tools show promise for automated analysis and modification of children’s speech production within assessment and therapeutic applications. Further work is needed to train automated systems with larger samples of speech to increase accuracy for assessment and therapeutic feedback.