No description available.

No description available.

The objective of this research is to develop a Markovian model in the form of a discrete-time queueing network to assess the performance of random access code division multiple access networks (CDMA). An approximation method called equilibrium point analysis (EPA) has been used to solve the model. The CDMA protocol IS an important application of spread spectrum communications that allows simultaneous transmission of multiple users to occupy a wideband channel with small interference. This is done by assigning each user a unique pseudo noise code sequence. These codes have low cross-correlation between each pair of sequences. Both slotted direct sequence CDMA (DS) and frequency hopping CDMA (FH) are considered with an emphasis on DS-CDMA systems. The EPA method has previously been used to evaluate the performance of other random access systems such as the ALOHA protocol, but has not previously been used in the context of a CDMA protocol. Throughput and mean packet delay of random access CDMA networks are evaluated, since these two measures are usually used in the study of the performance assessment of mUltiple access protocols. The analytical results of the random access model are validated against a discrete-event simulation which is run for large number of slots. The study then proceeds by using the model to examine the effect on performance of introducing error correcting codes to the DS-CDMA systems. Optimum error correcting codes that give the best performances in terms of the throughput and the delay are determined. The perfonnance of random access CDMA systems applied to radio channels, as in packet radio networks, is then studied, and the effect of multipath fading on the perfonnance is evaluated. Finally, the perfonnance of DS-CDMA with different user classes (non-identical users case) is investigated. An extended equilibrium point analysis (EEPA) method has been used to solve the Markovian model in this situation. This extended model is used to assess the effects on perfonnance of the unequal powers due to varying distances of the users to an intended receiver or to a base station (near-far problem).

Hypertonia is seen in 85% of people with Multiple Sclerosis (pwMS) resulting in disability and functional restrictions. Hypertonia can be caused by increases in passive stiffness and enhanced stretch reflexes (spasticity) and is frequently managed clinically using passive stretches. However, the optimal parameters of stretching such as the applied torque and stretch duration remain unclear. During commonly prescribed ankle plantarflexor stretches pwMS produced higher torques when standing in a weight bearing position compared to stretches applied using the upper limbs. Stretches could be held for 120 seconds on average and stretch duration was mainly limited by fatigue. People with higher disability tended to favour more supported stretching positions. The effects of stretching for either 30 or 10 minutes using a customised motor at three torque levels covering the range that MS participants could produce was investigated. Compared to the 10 minute stretch, greater reductions in passive stiffness and greater increases in range of movement (ROM) were seen immediately following the 30 minute stretch with the effects being sustained for the 30 minute post stretch period. Higher levels of applied torque resulted in a greater change in ROM however; there was no effect of applied torque on passive stiffness. Stretch reflex mediated stiffness was unaffected by the stretching intervention and showed transient post stretch increases. Ultrasonography was used to investigate changes in muscle–tendon length and strain in pwMS and controls and following stretching. PwMS showed evidence of stiffer muscles and increased tendon length at baseline compared to controls. Following a 10 minute stretch overall muscle length did not increase in pwMS, although increases in strain in the musculotendinous junction region were observed suggesting that more proximal regions of the muscle was likely to have contributed significantly to overall stiffness. This work highlights that stretch duration and levels of applied torque are critical factors in determining the effectiveness of stretches. The pathological mechanisms underlying hypertonia at a molecular and structural level and the effects of stretching on components of the musculo-tendinous structure and on functional ability should be ascertained.

This thesis consists of three essays on optimal monetary policy. In the first essay I study time-consistent monetary policy in an small open economy model with incomplete financial markets. I demonstrate the existence of two discretionary equilibria. The model is capable of explaining periods of different exchange rate volatilities as well as the transition between those regimes. Following a shock the economy can be stabilised either `quickly' or `slow', where both dynamic paths satisfy the conditions of optimality and time-consistency. I also show that a policy of partially targeting the exchange rate results in far worse welfare outcomes relative to a strict inflation targeting policy. In the second essay, I analyse how a policy maker can avoid expectation traps and coordination failures. Using a framework developed by Schaumburg and Tambalotti (2007) and Debortoli and Nunes (2010) in which a policy maker may or may not default on past promises I show that already mild degrees of precommitment are sufficient to generate uniqueness of the Pareto-preferred equilibrium. In the last chapter, I examine optimal monetary policy from an empirical perspective. I estimate a simple small open economy model separately for a policy maker acting under commitment and discretion and find that the data favours the commitment approach. Furthermore, the data suggest that the Bank of Canada did not target the nominal exchange rate in the inspected time period.

Several new multiuser detectors are developed for different direct-sequence codedivision multiple-access (DS-CDMA) application environments. The first detector is based on a semidefinite-programming (SDP) relaxation technique. In this detector, maximum likelihood (ML) detection is achieved by 'relaxing' the associated combinatorial problem into an SDP problem, which leads to a detector of polynomial complexity. It is shown that the SDP-relaxation (SDPR) based detector can be obtained by solving a dual SDP problem which leads to improved efficiency. Computer simulations demonstrate that the SDPR detector offers near-optimal performance with much reduced computational complexity compared with that of the ML detector proposed by Verdu for both synchronous and asynchronous DS-CDMA systems. The second detector is based on a recursive convex programming (RCP) approach. In this detector, ML detection is carried out in two steps: first, the combinatorial problem associated with ML detection is relaxed to a convex programming problem, and then a recursive approach is used to obtain an approximate solution for ML detection. Efficient unconstrained relaxation approach is proposed for the proposed detector to reduce the involved computational complexity. Computer simulations demonstrate that the proposed detectors offer near-optimal detection performance which is superior to that offered by many other suboptimal detectors including the SDPR detector. However, the computational complexity involved in the proposed detectors is much lower relative to that involved in Verdu's ML detector as well as our SDPR detector. The third detector entails a subspace estimation-based constrained optimization approach for channel estimation in DS-CDMA systems with multipath propagation channels. The proposed approach offers an improved approximation for the noise iii subspace compared with that offered by several existing algorithms. Computer simulations show that the performance of the proposed detector offers nearly the same performance as that of existing subspace detectors but leads to a significant reduction in the amount of computation. Relative to some existing constrained optimization methods, the proposed detector offers a significantly improved performance while requiring a comparable amount of computation. The fourth detector is proposed based on a vector constant-modulus (VCM) approach. This detector is designed for DS-CDMA systems with multipath propagation channels where the effective signatures observed at receiver are distorted by multipath propagation and aliasing concurrently. In this detector, detection is carried out by solving a linear constrained optimization problem whose objective function is formulated based on the VCM criterion. Two adaptation algorithms, namely, the constrained stochastic gradient algorithm and the recursive vector constant-modulus algorithm, are developed. Analysis are presented to investigate the performance of the proposed detector. Computer simulations show that the proposed detectors are able to suppress multiuser interference and inter-symbol interference effectively. More importantly, they offer robust detection performance against the effective signature distortion caused by aliasing at the receiver.

In this dissertation, direct-sequence code-division multiple access (DS-CDMA) systems with multiuser detectors used at receiver are investigated and two kinds of multiuser detectors are developed for DS-CDMA systems. In the investigation of DS-CDMA systems using multiuser detectors at receiver, a study on the performance of the system is presented, where heterogeneous traffic with different transmission rates and quality of service (QoS) requirements is supported. The effects of some realistic factors, such as imperfect power control and the existence of multiple cells, on the system performance are studied. In addition, algorithms are proposed to deal with the forward link power allocation problem based on the measurements of random characteristics of the received signals. This power allocation problem is formulated as a constrained optimization problem. To make the problem easy to solve, an additional appropriate constraint is proposed. Two methods are developed to identify the feasible region of this constrained optimization problem. The first proposed multiuser detector is an adaptive minimum mean-squared-error (MMSE) detector. Particularly, it is desirable for the cases where communication channels have severe near-far problem, and thus the convergence rates of adaptive MMSE detectors for users with different power are quite different. To improve the convergence rates of adaptive MMSE detectors for weak power users, the interference effects of the strong power user signals are subtracted from the received signal successively. The method to estimate the parameters required in the proposed detector is also developed. It is shown that the proposed detector achieves fast convergence rates in various near-far scenarios. Other studies conducted include the transient mean-squared-error (MSE) analysis to explain the different convergence rates of adaptive MMSE detectors for users with different power, and the bit-error-rate (BER) performance analysis for the proposed detector. The second proposed multiuser detector is a set of semi-blind linear parallel interference cancellation (PIC) detectors for the reverse link of multiple-cell systems, where only information about intra-cell users is available. To decrease the interference from inter-cell users whose information is unavailable to the receiver, the inter-cell user signal subspace is identified first by making use of the available information about intra-cell users. The eigenvectors and eigenvalues of this signal subspace are then used in the traditional linear PIC structure, in place of the unknown inter-cell users’ signature codes and signal amplitudes. Based on this idea, three detection schemes are proposed. In addition, an efficient adaptation implementation method is developed, and the performance of the proposed detectors is studied. The proposed detectors are shown to be suitable for practical implementations and have satisfactory performance. / Graduate

Deep learning networks in the literature traditionally only used a single input modality (or data stream). Integrating multiple modalities into deep learning networks with the goal of correlating extracted features was a major issue. Traditional methods involved treating each modality separately and then writing custom code to combine the extracted features. Current solutions for small numbers of modalities (three or less) showed there are multiple architectures for modality integration. With an increase in the number of modalities, the “curse of dimensionality” affects the performance of the system. The research showed current methods for larger scale integrations required separate, custom created modules with another integration layer outside the deep learning network. These current solutions do not scale well nor provide good generalized performance. This research report studied architectures using multiple modalities and the creation of a scalable and efficient architecture.

University of Macau / Faculty of Social Sciences / Department of Sociology

No description available.