Khalid Abualsaud

Ensemble classifier for epileptic seizure detection for imperfect EEG data

Brain status information is captured by physiological electroencephalogram (EEG) signals, which are extensively used to study different brain activities. This study investigates the use of a new ensemble classifier to detect an epileptic seizure from compressed and noisy EEG signals. This noise-aware signal combination (NSC) ensemble classifier combines four classification models based on their individual performance. The main objective of the proposed classifier is to enhance the classification accuracy in the presence of noisy and incomplete information while preserving a reasonable amount of complexity. The experimental results show the effectiveness of the NSC technique, which yields higher accuracies of 90% for noiseless data compared with 85%, 85.9%, and 89.5% in other experiments. The accuracy for the proposed method is 80% when SNR = 1 dB, 84% when SNR = 5 dB, and 88% when SNR = 10 dB, while the compression ratio (CR) is 85.35% for all of the datasets mentioned.

Performance evaluation for compression-accuracy trade-off using compressive sensing for EEG-based epileptic seizure detection in wireless tele-monitoring

Brain is the most important part in the human body controlling muscles and nerves; Electroencephalogram (EEG) signals record brain electric activities. EEG signals capture important information pertinent to different physiological brain states. In this paper, we propose an efficient framework for evaluating the power-accuracy trade-off for EEG-based compressive sensing and classification techniques in the context of epileptic seizure detection in wireless tele-monitoring. The framework incorporates compressive sensing-based energy-efficient compression, and noisy wireless communication channel to study the effect on the application accuracy. Discrete cosine transform (DCT) and compressive sensing are used for EEG signals acquisition and compression. To obtain low-complexity energy-efficient, the best data accuracy with higher compression ratio is sought. A reconstructed algorithm derived from DCT of daubechies wavelet 6 is used to decompose the EEG signal at different levels. DCT is combined with the best basis function neural networks for EEG signals classification. Extensive experimental work is conducted, utilizing four classification models. The obtained results show an improvement in classification accuracies and an optimal classification rate of about 95% is achieved when using NN classifier at 85% of CR in the case of no SNR value. The satisfying results demonstrate the effect of efficient compression on maximizing the sensor lifetime without affecting the applications accuracy.

Walsh transform with moving average filtering for data compression in wireless sensor networks

Due to the peculiarity of wireless sensor networks (WSNs), where a group of sensors continuously transmit data to other sensors or to the fusion center, it is crucial to compress the transmitted data in order to save the consumed power, which is paramount in the case of portable devices. There exists several techniques for data compression such as discrete wavelet transform (DWT) based, which fails to achieve high compression ratio for an acceptable distortion ratio. In this paper, we explore exploiting Walsh transform with a moving average filtering (MAF) for data compression in WSNs. One application of WSN is wireless body sensor networks. We apply Walsh transform on real Electroencephalogram (EEG) data collected from patients. Furthermore, we compare our results to DWT and show the superiority of exploiting Walsh transform for data compression. We show that using MAF with Walsh transform enhances the compression ratio for up to 30% more than that achieved by DWT.

Performance evaluation of experimental damage detection in structure health monitoring using acceleration

Wireless sensor networks (WSNs) are one of the emerging technologies in the 21st century. In the structural health monitoring (SHM). WSNs are used as one of the vitally capable technologies in the SHM. The accelerometer module in the existing sensor nodes enables several novel applications. In this paper, a prototype for monitoring and detecting the damage for the real bridge using these sensor nodes is built. The prototype consists of sensor nodes, shaking table including its amplifier, and real bridge. The sensors are placed on a scaled down concrete bridge model that is mounted on a shaking table. The results are demonstrated in terms of acceleration on different nodes at a particular excitation frequency in the case of normal, single-side damage, and double-side damage.

FPGA implementation of DWT EEG data compression for wireless body sensor networks

Wireless body sensor networks (WBSN) provide an appreciable aid to patients who require continuous care and monitoring. One key application of WBSN is mobile health (mHealth) for continuous patient monitoring, acquiring vital signs e.g. EEG, ECG, etc. Such monitoring devices are doomed to be portable, i.e., batter powered, and agile to allow for patient mobility, while providing sustainable, energy-efficient hardware platforms. Hence, EEG data compression is critical in reducing the transmission power, hence increase the battery life. In this paper, we design and implement a complete hardware model based on discrete wavelet transform (DWT) for vital signs data compression and reconstruction on a field programmable gate array (FPGA) based platform. We evaluate the performance of our DWT compression FPGA implementation under different practical parameters including filter length and the compression ratio. We investigate the hardware and computational complexity of our design in terms of used resource blocks for future comparison with state-of-the-art techniques. Our results show the efficiency of the proposed hardware compression and reconstruction model at different system parameters, including the high pass filter coefficients, and DWT type, and DWT threshold.

A new WDM Application Response Time in WLAN Network and Fixed WiMAX using Distributed

Worldwide Interoperability for Microwave Access (WiMAX) and Wireless LAN (WLAN) has emerged as a promising solution for last mile access technology to provide high speed internet access in the residential as well as small and medium sized enterprise sectors. Application Response Time is the key performance measure in WiMAX and WLAN Network Quality of Service (QoS). The WiMAX network does not provide sufficient QoS with respect to Application Response Time. Wavelength Division Multiplexing (WDM) has emerged as the promising technology to meet the ever-increasing demand for bandwidth In this paper, we developed a Distributed Client-Server Model to improve QoS with respect to Application Response Time in the Fixed WiMAX and WLAN Network in order to enhance the services that are provided to the end users. The new distributed Client-Server model was simulated in OPNET modeler 16.0 with multiple Base Stations (BSs), Subscribers Stations (SSs) and some Server BSs selected by the Nearest Neighborhood Algorithms using Orthogonal Frequency Division Multiplexing (OFDM) techniques and compared with the existing Centralized model using Frequency Division Multiplexing (FDM) techniques. The simulation results obtained for the application response time of the proposed Client-Server model show an improvement in network performance.

Performance Comparison of classification algorithms for EEG-based remote epileptic seizure detection in Wireless Sensor Networks

Identification of epileptic seizure remotely by analyzing the electroencephalography (EEG) signal is very important for scalable sensor-based health systems. Classification is the most important technique for wide-ranging applications to categorize the items according to its features with respect to predefined set of classes. In this paper, we conduct a performance evaluation based on the noiseless and noisy EEG-based epileptic seizure data using various classification algorithms including BayesNet, DecisionTable, IBK, J48/C4.5, and VFI. The reconstructed and noisy EEG data are decomposed with discrete cosine transform into several sub-bands. In addition, some of statistical features are extracted from the wavelet coefficients to represent the whole EEG data inputs into the classifiers. Benchmark on widely used dataset is utilized for automatic epileptic seizure detection including both normal and epileptic EEG datasets. The classification accuracy results confirm that the selected classifiers have greater potentiality to identify the noisy epileptic disorders.