No abstract

In this paper, the performance of five classifiers in P300 speller paradigm are compared. Theses classifiers are Linear Support Vector Machine (LSVM), Gaussian Support Vector Machine (GSVM), Neural Network (NN), Fisher Linear Discriminant (FLD), and Kernel Fisher Discriminant (KFD). In classification of P300 waves, there has been a trend to use SVM classifiers. Although they have shown a good performance, in this paper, it is shown that the FLD classifiers outperform the SVM classifiers. FLD classifier uses only ten channels of the recorded electroencephalogram (EEG) signals. This makes them a very good candidate for real-time applications. In addition, FLD approach does not need any optimization similar to other methods. In addition, in this paper, it is shown that the efficiency of using Principal Component Analysis (PCA) for feature reduction results in decreasing the time for the classification and increasing the accuracy.

Fog computing is a potential solution for heterogeneous resource-constrained mobile devices to collaboratively operate deep learning-driven applications at the edge of the networks, instead of offloading the computations of these applications to the powerful cloud servers thanks to the latency reduction, decentralized structure, and privacy concerns. Compared to the mobile cloud computing concept where computation-intensive deep learning operations are offloaded to the powerful cloud servers, making use of the computing capabilities of resource-constrained devices can improve the delay performance and lessen the need for powerful servers to execute such applications by considering a collaborative fog computing scenario with deep neural network (DNN) partitioning. In this paper, we propose an energy-efficient finegrained DNN partitioning scheme for wireless collaborative fog computing systems. The proposed scheme includes both layer-based partitioning where the DNN model is divided into layer by layer and horizontal partitioning where the input data of each layer operation is partitioned among multiple devices to encourage parallel computing. A convex optimization problem is formulated to minimize the energy consumption of the collaborative part of the system by optimizing the communication and computation parameters as well as the workload of each participating device and solved by using the primal-dual decomposition and Lagrange duality theory. As can be observed in the simulation results, the proposed optimized scheme makes a notable difference in the energy consumption compared to the non-optimized scenario where the workload distribution is equal for all participating devices but the communication and computation parameters are still optimized, so it is a quite challenging bound to be compared.INDEX TERMS Convex optimization, deep convolutional neural network, energy efficiency, fog computing, DNN partitioning, wireless collaborative computing.

In this work, a heterogeneous set of wireless devices sharing a common access point (AP) or base station (BS) collaborates to complete a set of computing tasks within a given deadline in the most energy-efficient way. This pool of devices somehow acts like a distributed mobile edge computing (MEC) server to augment the computing capabilities of individual devices while reducing their total energy consumption. Using the Map-Reduce distributed computing framework -which involves both local computing at devices and communications between them -the tasks are optimally distributed amongst the nodes, taking into account their diversity in term of computing and communications capabilities.In addition to optimizing the computing load distribution, local parameters of the nodes such as CPU frequency and RF transmit power are also optimized for energy-efficiency. The corresponding optimization problem can be shown to be convex and optimality conditions offering insights into the structure of the solutions can be obtained through Lagrange duality theory. A waterfilling-like interpretation for the size of the computing task assigned to each node is given. Numerical experiments demonstrate the benefits of the proposed optimal collaborative-computing scheme over various other schemes in several respects.Most notably, the proposed scheme exhibits increased probability of successfully dealing with larger computing loads and/or smaller latency and energy-efficiency gains of up to two orders of magnitude.Both improvements come from the scheme ability to optimally leverage devices diversity.

Artifact removal is an essential part in electroencephalogram (EEG) recording and the raw EEG signals require preprocessing before feature extraction. In this work, we implemented three filtering methods and demonstrated their effects on the performance of different classifiers. Bandpass digital filtering, median filtering and facet method are three preprocessing approaches investigated in this paper. We used data set lib from the BCI competition 2003 for training and testing phase. Our accuracy varied between 80% and 96%. In our work, we demonstrated that the problems of choosing the classifier and preprocessing methods are not independent of each other. Two of our approaches could achieve the 96% accuracy i.e. 31 of 32 characters were predicted correctly. These two approaches have different classifier and different preprocessing method. It means that the performance of each classifier can be enhanced with a specific preprocessing method. In our approach, we used only three electrodes of 64 applied electrodes. Therefore it can noticeably reduce the time and cost of EEG measurement.