A5039591248- Speech and Audio Processing
- Advanced Adaptive Filtering Techniques
- Hearing Loss and Rehabilitation
- ECG Monitoring and Analysis
- Acoustic Wave Phenomena Research
- Blind Source Separation Techniques
- Speech Recognition and Synthesis
- Heart Rate Variability and Autonomic Control
- Non-Invasive Vital Sign Monitoring
- EEG and Brain-Computer Interfaces
- Dysphagia Assessment and Management
- Tracheal and airway disorders
- Music and Audio Processing
- Human Pose and Action Recognition
- Ultrasonics and Acoustic Wave Propagation
- Phonocardiography and Auscultation Techniques
- Image Retrieval and Classification Techniques
- Anomaly Detection Techniques and Applications
- Analog and Mixed-Signal Circuit Design
- Advanced Neural Network Applications
- Infant Development and Preterm Care
- Sleep and Work-Related Fatigue
- Face and Expression Recognition
- Online and Blended Learning
- Mathematics Education and Programs
Griffith University
2012-2025
Abstract Cervical auscultation, commonly used by speech-language pathologists in some countries as an adjuvant to the clinical feeding evaluation, requires data on acoustic and perceptual profiles of swallowing sounds. Whilst these exists adults children, none currently exist for preterm neonates. Our study aims establish parameters sounds Swallowing were recorded a digital microphone during oral observations. Acoustic duration, peak frequency, power intensity determined. Perceptual heard...
Abstract Use of machine learning to accurately detect aspirating swallowing sounds in children is an evolving field. Previously reported classifiers for the detection have sensitivities between 79 and 89%. This study aimed investigate accuracy using automatic speaker recognition approach differentiate normal recorded from digital cervical auscultation children. We analysed 106 swallows 23 healthy (median 13 months; 52.1% male) 18 10.5 61.1% who underwent concurrent videofluoroscopic swallow...
This paper presents the results of a project investigating use discrete-cosine transform (DCT) to represent facial images an identity recognition system. The objective was address problem sensitivity variation in illumination, faced by commercially available systems. proposed method uses local, block-wise DCT extract features image. Results show that this appearance based gives high identification rates, improved tolerance and simple implementation.
In this work, we propose a novel algorithm to achieve real-time R-peak prediction during ECG signal recording. More specifically, from the current frame of signal, aim predict how far into future next will occur, taking account information about variability in intervals between beats seen previous frames signal. Currently there has been little work area. However, beat important research significance, including timing artificial heart pumps and integration cardiopulmonary support (CPS-heart),...
In this paper we investigate the enhancement of speech by applying MMSE short-time spectral magnitude estimation in modulation domain. For purpose, traditional analysismodification-synthesis framework is extended to include domain processing. We compensate noisy spectrum for additive noise distortion algorithm Subjective experiments were conducted compare quality stimuli processed estimator those using acoustic and subtraction method. The proposed method shown have better suppression than...
A number of techniques based on correlation measurements have recently been proposed to provide an objective measure intelligibility. These are able detect nonlinear distortions and intelligibility scores highly correlated with those given by human listeners. However, the performance these has not found satisfactory for measuring speech enhancement algorithms. In this paper we first investigate different correlation-based methods, in context enhancement. We then propose combine spectral...
Electrocardiography (ECG) is a promising approach for continuous fetal heart rate monitoring. Its morphology can provide information on health to guide patient care by clinicians. However, ECGs extracted from abdominal are often too weak reliably detect rate. This study evaluates the application of U-Net architecture accurate R-peak detection in low-SNR ECG signals. The proposed method achieves high accuracy with positive predictive value 99.81%, sensitivity 100.00%, and an F1-score 99.91%...
In this paper we investigate the modulation domain as an alternative to acoustic for speech enhancement. More specifically, wish determine how competitive is spectral subtraction compared domain. For purpose, extend traditional analysis-modification-synthesis framework include processing. We then compensate noisy spectrum additive noise distortion by applying algorithm in Using subjective listening tests and objective quality evaluation show that proposed method results improved quality....
With the current day advancements in both computational power and machine learning (ML) techniques, there is a fundamental shift toward application of new smarter technologies. Worldwide incidents motor vehicle crashes cause financial emotional distress, along with physical injury, even death, often stemming from driver fatigue. Nowadays, advanced ML techniques can be combined electrocardiogram signals recorded hand-contact steering wheel, to accurately detect onset However, signal only...
In this paper, we investigate the advantages of applying subspace approach within short-time Fourier transform-based (STFT) modulation domains over conventional sub-space algorithm for single-channel speech enhancement. Specifically, consider processing modulations STFT real and imaginary (RI) parts magnitude values. Speech enhancement experiments were performed on Noizeus corpus that had been corrupted by varying levels white coloured noise. The results indicate domain-based methods...
In this paper, we present a new speech enhancement method that processes noise-corrupted in the discrete cosine transform (DCT) modulation domain. contrast to Fourier transform, DCT produces real-valued signal. Therefore, modulation-based processing domain may allow both acoustic magnitude and phase information be jointly estimated. Based on segmental SNR results of blind subjective listening tests corrupted with various coloured noises, application subspace was found outperform all other...
Analysis and assessment of human movement are crucial for enhancing athletic performance, preventing injuries, aiding rehabilitation. This paper applies the most recent iteration You Only Look Once (YOLO) family algorithms (version 8) to resistance training, segmenting humans gym equipment in real time. Velocity metrics calculated directly from segmentation masks, adding growing body research concerned with velocity-based training (VBT), an emerging alternative traditional percentage-based...