Search

The importance of speech enhancement (SE) is increased by its numerous real-life applications in hearing aids, speech recognition, speech coding, and cochlear implants. Transform-based speech enhancement algorithm (SEA), in particular, is the most popular, as the noise energy can be more easily distinguished and removed from the speech signal in the transform domain. Most speech enhancement systems use discrete Fourier transform (DFT), which has readily available short-time spectral amplitude (STSA) estimators. There are limitations with regard to the conventional DFT-based STSA estimators. For example, using a noisy phase spectrum for speech re-synthesis introduces an upper bound on the maximum improvement in speech quality; also, no close-form solution is available in the DFT domain when super-Gaussian speech prior is used for derivation. The approach presented here alleviates such limitations by using discrete cosine transform (DCT) instead of DFT, thereby improving perceived speech quality. Therefore, this thesis focuses on demonstrating the superiority of DCT representation in the context of STSA-estimation-based speech enhancement. As a foundation of the research, we first evaluate the relevance of DCT spectra towards perceived speech quality using objective and subjective testing. Consequently, the use of noisy polarity spectrum (PoS) for speech re-synthesis is justified. In the second stage of the research, we develop the optimal minimum mean square error (MMSE) estimators of the DCT STSA due to their primary importance on perceived speech quality. A novel polarity estimator (PoE) is also developed and used with the STSA estimator. Accordingly, we examine the effect of using PoE on the performance of the proposed SE system. Compared to the state-of-the-art DFT based STSA estimators, the proposed DCT-based approach demonstrates superior performance in enhancing noisy speech as indicated by objective and subjective quality measures., Thesis (PhD Doctorate), Doctor of Philosophy (PhD), School of Eng & Built Env, Science, Environment, Engineering and Technology, Full Text

Proteins are the most abundant macromolecules and essential to all life forms playing critical roles in many biological processes ranging from catalyst in enzymic reactions, antibodies counteracting antigens, transmitters of cellular signals, to acting as the structural elements in cells and tissues, and many more. The facet of proteins providing them with this functionality is their three-dimensional (3D) structure. Therefore, knowing the 3D structure of the protein is essential to understand its functionality and role. Recently, Alphafold2 has achieved atomic-level accuracy in predicting the 3D structure of the protein. This scientific achievement is built on accumulating improvement in protein one-dimensional (1D) and two-dimensional (2D) structural properties prediction along with advances in deep learning techniques. However, this success only partially solves the problem of protein structure prediction as it requires a minimum of 30 homologous sequences for effective prediction, and a large number of proteins lack homologous sequences. Consequently, it is still essential to improve the accuracy of protein secondary structure and inter-residue contact prediction for all proteins, specifically those without homologous sequences. This thesis explores several deep learning approaches, input features and improved training strategies for the prediction of protein secondary structure, 1D structural properties and inter-residues contact map prediction. It demonstrates leveraging large high sequence identity training sets and protein language models as input features enhance the prediction for all test sets, including those proteins which lack homologous sequences. There are mainly four research studies performed in this thesis., Thesis (PhD Doctorate), Doctor of Philosophy (PhD), School of Eng & Built Env, Science, Environment, Engineering and Technology, Full Text

Hyperspectral object tracking has recently emerged as a topic of great interest in the remote sensing community. The hyperspectral image, with its many bands, provides a rich source of material information of an object that can be effectively used for object tracking. While most hyperspectral trackers are based on detection-based techniques, no one has yet attempted to employ YOLO for detecting and tracking the object. This is due to the presence of multiple spectral bands, the scarcity of annotated hyperspectral videos, and YOLO's performance limitation in managing occlusions, and distinguishing object in cluttered backgrounds. Therefore, in this paper, we propose a novel framework called Hy-Tracker, which aims to bridge the gap between hyperspectral data and state-of-the-art object detection methods to leverage the strengths of YOLOv7 for object tracking in hyperspectral videos. Hy-Tracker not only introduces YOLOv7 but also innovatively incorporates a refined tracking module on top of YOLOv7. The tracker refines the initial detections produced by YOLOv7, leading to improved object-tracking performance. Furthermore, we incorporate Kalman-Filter into the tracker, which addresses the challenges posed by scale variation and occlusion. The experimental results on hyperspectral benchmark datasets demonstrate the effectiveness of Hy-Tracker in accurately tracking objects across frames.

Full Text, Thesis (PhD Doctorate), Doctor of Philosophy (PhD), School of Eng & Built Env, Science, Environment, Engineering and Technology, RNAs are important biological macro-molecules that play critical roles in many biological processes. The functionality of RNA depends on its three-dimensional (3D) structure, which further depends on its primary structure, i.e. the order of sequence of nucleotides in the RNA chain. Direct prediction of the 3D structure of an RNA from its sequence is a challenging task. Therefore, the 3D structure is further divided into two-dimensional (2D) properties such as secondary structure, contact maps and one-dimensional (1D) properties such as torsion angles and solvent accessibility. An accurate prediction of these 1D and 2D structural properties will increase the accuracy in predicting the 3D structure of the RNA. This thesis explores various deep learning algorithms and input features relevant to predicting the 1D and 2D structural properties of an RNA. Using these predicted 1D and 2D structural properties further as restraints, we have demonstrated an improvement in the prediction of the RNA 3D structure. There are four primary studies performed in this thesis for RNA structural properties prediction. The first study introduces two methods (SPOT-RNA and SPOT-RNA2) for RNA secondary structure prediction using an ensemble of Residual Con-volution and Bi-directional LSTM recurrent neural networks. This study shows that deep learning based methods can outperform existing dynamic programming based algorithms and achieve state-of-the-art performance using single-sequence and evolutionary information as input. The second study investigates the application of deep neural networks for predicting RNA backbone torsion and pseudotorsion angles. We have pioneered in predicting the backbone torsion and pseudotorsion angles using deep learning (SPOT-RNA-1D). The angles predicted using SPOT-RNA-1D could be used as 3D model quality indicators. The third study introduces a method (SPOT-RNA-2D) to predict RNA distance-based contact maps using an ensemble of deep neural networks and improv

Full Text, Thesis (PhD Doctorate), Doctor of Philosophy (PhD), School of Eng & Built Env, Science, Environment, Engineering and Technology, Speech corrupted by background noise (or noisy speech) can reduce the efficiency of communication between man-man and man-machine. A speech enhancement algorithm (SEA) can be used to suppress the embedded background noise and increase the quality and intelligibility of noisy speech. Many applications, such as speech communication systems, hearing aid devices, and speech recognition systems, typically rely upon speech enhancement algorithms for robustness. This dissertation focuses on single-channel speech enhancement using Kalman filtering with machine learning methods. In Kalman filter (KF)-based speech enhancement, each clean speech frame is represented by an auto-regressive (AR) process, whose parameters comprise the linear prediction coefficients (LPCs) and prediction error variance. The LPC parameters and the additive noise variance are used to form the recursive equations of the KF. In augmented KF (AKF), both the clean speech and additive noise LPC parameters are incorporated into an augmented matrix to construct the recursive equations of AKF. Given a frame of noisy speech samples, the KF and AKF give a linear MMSE estimate of the clean speech samples using the recursive equations. Usually, the inaccurate estimates of the parameters introduce bias in the KF and AKF gain, leading to a degradation in speech enhancement performance. The research contributions in this dissertation can be grouped into three focus areas. In the first work, we propose an iterative KF (IT-KF) to offset the bias in KF gain for speech enhancement through utilizing the parameters in real-life noise conditions. In the second work, we jointly incorporate the robustness and sensitivity metrics to offset the bias in the KF and AKF gain - which address speech enhancement in real-life noise conditions. The third focus area consists of the deep neural network (DNN) and whitening filter assisted KF and AKF for speech enhancement. Specifically, DNN and whitening filter-based approaches utilize

Full Text, Thesis (PhD Doctorate), Doctor of Philosophy (PhD), School of Eng & Built Env, Science, Environment, Engineering and Technology, Time-Scale Modification (TSM) is a well-researched field and allows for time-domain manipulation of a signal without modifying the pitch or timbre. Many TSM methods have been presented, however quantitative results on the quality of these methods are rare, with most methods reporting informal listening tests. This is likely due to the timecommitment and cost of subjective testing. Additionally, an objective measure of quality has not yet been developed that is suitable for timescaled signals. This dissertation describes the design of e ective objective measures of quality for TSM. TSM methods are, generally, single channel algorithms that give poor results when applied to multi-channel signals, as the phase relationship between channels must be maintained. This dissertation proposes a method and additional variant for maintaining the phase relationship between channels and retaining the presence in the centre of the stereo signal. The method involves pre- and post-processing the signal, with the variant processing each frame for real-time suitability. Sum and di erence transformations of the stereo signal are used for TSM and result in a large improvement in stereo phase coherence, consequently maintaining the stereo field. The proposed method produces a highquality stereo output and greatly improves quality over the independent channel processing method. It also allows for simple implementation around all existing TSM frameworks. A modification to the Epoch-Synchronous Overlap-Add (ESOLA) TSM algorithm is proposed in this dissertation. The proposed method, Fuzzy Epoch-Synchronous Overlap-Add, improves on the previous ESOLA method through cross-correlation of time-smeared epochs before overlap-adding. This reduces distortion and artefacts while the speaker's fundamental frequency is stable, as well as reducing artefacts during pitch modulation. The proposed method is tested against well-known TSM algorithms. It is preferred over ESOLA and gives similar performance t

Full Text, Thesis (PhD Doctorate), Doctor of Philosophy (PhD), School of Eng & Built Env, Science, Environment, Engineering and Technology, Speech corrupted by background noise (or noisy speech) can cause misinterpretation and fatigue during phone and conference calls, and for hearing aid users. Noisy speech can also severely impact the performance of speech processing systems such as automatic speech recognition (ASR), automatic speaker verification (ASV), and automatic speaker identification (ASI) systems. Currently, deep learning approaches are employed in an end-to-end fashion to improve robustness. The target speech (or clean speech) is used as the training target or large noisy speech datasets are used to facilitate multi-condition training. In this dissertation, we propose competitive alternatives to the preceding approaches by updating two classic robust speech processing techniques using deep learning. The two techniques include minimum mean-square error (MMSE) and missing data approaches. An MMSE estimator aims to improve the perceived quality and intelligibility of noisy speech. This is accomplished by suppressing any background noise without distorting the speech. Prior to the introduction of deep learning, MMSE estimators were the standard speech enhancement approach. MMSE estimators require the accurate estimation of the a priori signal-to-noise ratio (SNR) to attain a high level of speech enhancement performance. However, current methods produce a priori SNR estimates with a large tracking delay and a considerable amount of bias. Hence, we propose a deep learning approach to a priori SNR estimation that is significantly more accurate than previous estimators, called Deep Xi. Through objective and subjective testing across multiple conditions, such as real-world non-stationary and coloured noise sources at multiple SNR levels, we show that Deep Xi allows MMSE estimators to produce the highest quality enhanced speech amongst all clean speech magnitude spectrum estimators. Missing data approaches improve robustness by performing inference only on noisy speech features that reliably represent