Showing papers by "Azura Che Soh published in 2020"

ON-OFF Body Ultra-Wideband (UWB) Antenna for Wireless Body Area Networks (WBAN): A Review

Abstract: Ultra-wideband (UWB) technology can offer broad capacity, short-range communications at a relatively low level of energy usage, which is very desirable for wireless body area networks (WBANs). The involvement of the human body in such a device poses immense difficulties for both the architecture of the wearable antenna and the broadcast model. Initially, the bonding between the wearable antenna and the human body should also be acknowledged in the early stages of the design, so that both the potentially degrading output of the antenna as a consequence of the body and the possibility of exposure for the body may be handled. Next, the transmission path in WBAN is affected by the constant activity of the human body, leading to the time-varying dispersion of electromagnetic waves. Few researchers were interested in this field, and some substantial progress has recently been considered. On the other hand, this paper covered both wearable and Non-wearable UWB antenna designs and applications with respect to their substrate characteristics. Finally, this review prospectively exposes the upgraded developments of (ON-OFF) body antennas in the area of wearable and Non-wearable UWB and their implementations in the WBAN device and aims to evaluate the latest design features that inspire the performance of the antennas.

TRIGA PUSPATI reactor: model analysis and accuracy

Abstract: There are many challenging issues with research reactor, such as time variation and uncertainty. Since its first criticality in 1982, the biggest changes in TRIGA PUSPATI Reactor system is the replacement of instrumentation and control console system from analogue to digital in 2013. Apart from providing methods of controlling the power reactor via the control rod movement, the Instrumentation and Control Console System also provides monitoring and display for all reactor parameters to protect the reactor from undue influences or abnormal circumstances. Meanwhile, the simulation model of the TRIGA PUSPATI Reactor system has been developed in the Simulink-MATLAB. The simulation model development is based on the research reactor mathematical representatives and the real plant parameters of TRIGA PUSPATI Reactor. However, the performance of this simulation model needs to be evaluated. Since there is no report or paper work found on the performance of the simulation model to represent the real system of RTP, the present study aims to carry out an analysis for more rigorous understanding of the TRIGA PUSPATI Reactor model simulation through validation and verification methods. After analysing the result, it was found that the simulation model has a good representation of a real plant.

Improvement of LMS adaptive noise canceller using uniform Poly-phase digital filter bank

Abstract: This paper presents the Least Mean Square (LMS) noise canceller using uniform poly-phase digital filter bank to improve the noise can-cellation process. Analysis filter bank is used to decompose the full-band distorted input signal into sub-band signals. Decomposition the full-band input distorted signal into sub-band signals based on the fact that the signal to noise ratio (S/N) is inversely proportional to the signal bandwidth. Each sub-band signal is fed to individual LMS algorithm to produce the optimal sub-band output. Synthesis filter bank is used to compose the optimal sub-band outputs to produce the final optimal full-band output. In this paper, m-band uniform Discrete Fourier Transform (DFT) digital filter bank has been used because its computational complexity is much smaller than the direct implementation of digital filter bank. The simulation results show that the proposed method provides the efficient performance with less and smooth error signal as compared to conventional LMS noise canceller.

Artificial neural network model and fuzzy logic control of dissolved oxygen in a bioreactor

Abstract: In a fermentation process, dissolved oxygen is the one of the key process variables that needs to be controlled because of the effect they have on the product quality. In a penicillin production, dissolved oxygen concentration influenced biomass concentration. In this paper, multilayer perceptron neural network (MLP) and Radial Basis Function (RBF) neural network is used in modeling penicillin fermentation process. Process data from an industrial scale fed-batch bioreactor is used in developing the models with dissolved oxygen and penicillin concentration as the outputs. RBF neural network model gives better accuracy than MLP neural network. The model is further used in fuzzy logic controller design to simulate control of dissolved oxygen by manipulation of aeration rate. Simulation result shows that the fuzzy logic controller can control the dissolved oxygen based on the given profile.

A Promising Wavelet Decomposition –NNARX Model to Predict Flood: Application to Kelantan River Flood

Abstract: Flood is a major disaster that happens around the world. It has caused many casualties and massive destruction of property. Estimating the chance of a flood occurring depends on several factors, such as rainfall, the structure and the flow rate of the river. This research used the neural network autoregressive exogenous input (NNARX) model to predict floods. One of the research challenges was to develop accurate models and improve the forecasting model. This research aimed to improve the performance of the neural network model for flood prediction. A new technique was proposed for modelling nonlinear data of flood forecasting using the wavelet decomposition-NNARX approach. This paper discusses the process of identifying the parameters involved to make a forecast as the rainfall value requires the flow rate of the river and its water level. The original data were processed by wavelet decomposition and filtered to generate a new set of data for the NNARX prediction model where the process can be compared. This research compared the performance of the wavelet and the non-wavelet NNARX model. Experimental results showed that the proposed approach had better performance testing results in relation to its counterpart in terms of hourly forecast, with the mean square error (MSE) of 2.0491e-4 m2 compared to 6.1642e -4 m 2 , respectively. The proposed approach was also studied for long-term forecast up to 5 years, where the obtained MSE was higher, i.e., 0.0016 m 2 .

Deep learning object detector using a combination of Convolutional Neural Network (CNN) architecture (MiniVGGNet) and classic object detection algorithm

Abstract: The object detection system is a computer technology related to image processing and computer vision that detects instances of semantic objects of a certain class in digital images and videos. The system consists of two main processes, which are classification and detection. Once an object instance has been classified and detected, it is possible to obtain further information, including recognizes the specific instance, track the object over an image sequence and extract further information about the object and the scene. This paper presented an analysis performance of deep learning object detector by combining a deep learning Convolutional Neural Network (CNN) for object classification and applies classic object detection algorithms to devise our own deep learning object detector. MiniVGGNet is an architecture network used to train an object classification, and the data used for this purpose was collected from specific indoor environment building. For object detection, sliding windows and image pyramids were used to localize and detect objects at different locations, and non-maxima suppression (NMS) was used to obtain the final bounding box to localize the object location. Based on the experiment result, the percentage of classification accuracy of the network is 80% to 90% and the time for the system to detect the object is less than 15sec/frame. Experimental results show that there are reasonable and efficient to combine classic object detection method with a deep learning classification approach. The performance of this method can work in some specific use cases and effectively solving the problem of the inaccurate classification and detection of typical features.

Development of fall detection and activity recognition using threshold based method and neural network

Abstract: Falls are dangerous and contribute to over 80% of injury-related hospitalization especially amongst the elderly. Hence, fall detection is important for preventing severe injuries and accidental deaths. Meanwhile, recognizing human activity is important for monitoring health status and quality of life as it can be applied in geriatric care and healthcare in general. This research presents the development of a fall detection and human activity recognition system using Threshold Based Method (TBM) and Neural Network (NN). Intentional forward fall and six other activities of daily living (ADLs), which include running, jumping, walking, sitting, lying, and standing are performed by 15 healthy volunteers in a series of experiments. There are four important stages involved in fall detection and ADL recognition, which are signal filtering, segmentation, features extraction and classification. For classification, TBM achieved an accuracy of 98.41% and 95.40% for fall detection and activity recognition respectively whereas NN achieved an accuracy of 97.78% and 96.77% for fall detection and activity recognition respectively.

Simulated Real-Time Controller for Tuning Algorithm Using Modified Hill Climbing Approach Based on Model Reference Adaptive Control System

Abstract: Many techniques and inventions in the field of automatic control keeps going forwards, especially that the introduction of modern computing provided a huge ground for innovation in this field. Intelligent control was one of the direct beneficiaries of the computer systems advancements. That led to new ways to tackle old problems like model inaccuracies and inconsistencies. Often, it is necessary to calibrate a certain parameters of a control system due to plant parameters fluctuation over time.In this research, an intelligent algorithmic tuning technique suitable for realtime system tuning based on hill climbing optimization algorithm and model reference adaptive control system (MRAC) technique is proposed. Although all adaptive control tuning methodologies depend partially or completely on online plant system identification, the proposed method uses only the model that is used to design the original controller, leading to simplified calculations that require neither high processing power nor long processing time, as opposed to identification techniques calculations. The main principle in the tuning process is to compare the output of the plant with a desired reference signal within an acceptable error margin. In order to investigate the ability of the proposed tuning method to deal with different system complexities, simulations of three different case studies were conducted. In each case study, different possibilities to generate the desired reference signal is discussed along with how much the complexity of the system would affect the end result. Also, in each case study a discussion contrasts the limitations and conditions needed to be identified to use the proposed method. The proposed design performed very well, improving overshoot and response speeds in example systems depending on reference response generation method. The results showed that using different methods to generate the reference response gives system designer flexibility over favouring a specific response characteristics or an overall decent response. The simulation results illustrates that the method schemes proposed in this study show a viable and versatile solution to deal with controller tuning for systems with model inaccuracies as well as controller real time calibration problem.

Linear regression technique for improvement of feet position estimation during standing balance using a cane with millimeter wave radar

Abstract: It is estimated that one in three seniors fall at least once a year. Falls are a global problem for the elderly that affects their quality of life and poses a great risk. In our research, we are trying to develop a system that could prevent falls by estimating the fall risk in real time. The system would measure the balance of the user by measuring the position of the Center of Gravity inside the Base of Support. In our previous research, we presented a system with a millimeter wave radar attached to a cane to measure the area of the Base of Support. However, the obtained results for the foot position estimation error were significantly worse than similar studies. One of the reasons was that the sensor was not really estimating the position of the feet but the position of the lower legs. Therefore, in this research we present a correction model to improve the feet position estimation. The proposed model was able to reduce the foot position estimation RMSE from 54 mm down to 34 mm, which is closer to the results of other similar studies measuring the position of the feet.