Directed Graph-based Wireless EEG Sensor Channel Selection Approach for Cognitive Task Classification

📝 Abstract

Wireless electroencephalogram (EEG) sensors have been successfully applied in many medical and computer brain interface classifications. A common characteristic of wireless EEG sensors is that they are low powered devices, and hence an efficient usage of sensor energy resources is critical for any practical application. One way of minimizing energy consumption by the EEG sensors is by reducing the number of EEG channels participating in the classification process. For the purpose of classifying EEG signals, we propose a directed acyclic graph (DAG)-based channel selection algorithm. To achieve this objective, the EEG sensor channels are first realized in a complete undirected graph, where each channel is represented by a node. An edge between any two nodes indicates the collaboration between these nodes in identifying the system state; and the significance of this collaboration is quantified by a weight assigned to the edge. The complete graph is then reduced into a directed acyclic graph that encodes the knowledge of the non-increasing order of the channel ranking for each cognitive task. The channel selection algorithm utilizes this directed graph to find a maximum path such that the total weight of this path satisfies a predefined threshold. It has been demonstrated experimentally that channel utilization has been reduced by 50% in the worst case scenario for a three-state system and an EEG sensor with 14 channels; and the best classification accuracy obtained is 81%.

💡 Analysis

Wireless electroencephalogram (EEG) sensors have been successfully applied in many medical and computer brain interface classifications. A common characteristic of wireless EEG sensors is that they are low powered devices, and hence an efficient usage of sensor energy resources is critical for any practical application. One way of minimizing energy consumption by the EEG sensors is by reducing the number of EEG channels participating in the classification process. For the purpose of classifying EEG signals, we propose a directed acyclic graph (DAG)-based channel selection algorithm. To achieve this objective, the EEG sensor channels are first realized in a complete undirected graph, where each channel is represented by a node. An edge between any two nodes indicates the collaboration between these nodes in identifying the system state; and the significance of this collaboration is quantified by a weight assigned to the edge. The complete graph is then reduced into a directed acyclic graph that encodes the knowledge of the non-increasing order of the channel ranking for each cognitive task. The channel selection algorithm utilizes this directed graph to find a maximum path such that the total weight of this path satisfies a predefined threshold. It has been demonstrated experimentally that channel utilization has been reduced by 50% in the worst case scenario for a three-state system and an EEG sensor with 14 channels; and the best classification accuracy obtained is 81%.

📄 Content

Directed Graph-based Wireless EEG Sensor Channel Selection Approach for Cognitive Task Classification
Abduljalil Mohamed*, Khaled Bashir Shaban**, and Amr Mohamed** *Information Systems Department, Ahmed Bin Mohamed Military College P.O. Box: 22988, Doha, Qatar **Computer Science and Engineering Department, College of Engineering, Qatar University P.O. Box: 2713, Doha, Qatar ajamoham@abmmc.edu.qa, khaled.shaban@qu.edu.qa, amrm@qu.edu.qa

Abstract—Wireless electroencephalogram (EEG) sensors have been successfully applied in many medical and computer brain interface classifications. A common characteristic of wireless EEG sensors is that they are low powered devices, and hence an efficient usage of sensor energy resources is critical for any practical application. One way of minimizing energy consumption by the EEG sensors is by reducing the number of EEG channels participating in the classification process. For the purpose of classifying EEG signals, we propose a directed acyclic graph (DAG)-based channel selection algorithm. To achieve this objective, the EEG sensor channels are first realized in a complete undirected graph, where each channel is represented by a node. An edge between any two nodes indicates the collaboration between these nodes in identifying the system state; and the significance of this collaboration is quantified by a weight assigned to the edge. The complete graph is then reduced into a directed acyclic graph that encodes the knowledge of the non-increasing order of the channel ranking for each cognitive task. The channel selection algorithm utilizes this directed graph to find a maximum path such that the total weight of this path satisfies a predefined threshold. It has been demonstrated experimentally that channel utilization has been reduced by 50% in the worst case scenario for a three-state system and an EEG sensor with 14 channels; and the best classification accuracy obtained is 81%.
Keywords- Brain computer interface; channel selection; graph theory; wireless EEG signal classification. I. INTRODUCTION
Recent advances in wireless communication and intelligent wireless electroencephalogram (EEG) sensors have allowed the realization of different EEG-based applications, such as healthcare [1] and brain computer interface applications [2]. In general, an EEG sensor network topology comprises of simple sensors that collect information about the subject and send it through wireless paths to the sink. A common characteristic of the wireless EEG sensors is that they are low powered devices, and hence an efficient usage of sensor energy resources is critical for any practical application. As shown in [3, 4], the sensing and processing energy are negligible with respect to communication energy. Thus, most of the energy-aware algorithms reported in the literature address this issue at the communication level [5- 7]. It is also shown in [8] that the reduction of the number of sensors can also reduce the power consumption for wireless EEG caps. Moreover, it yields more comfort for the user, decreases installation time duration and may substantially reduce the financial cost of the Brain Computer Interface (BCI) setup since the cost of an EEG cap and an amplifier vary in relation to the number of channels [9].
In this paper we address the problem of channel reduction for multi-channel sensor acquiring EEG data for a BCI classifier identifying cognitive tasks. The channel selection problem can be well represented as a complete undirected graph, where each channel is represented as a graph node. An edge between any two nodes indicates the collaboration between the corresponding channels in identifying the current system state; and the significance of this collaboration is quantified by a weight assigned to the edge.
Various approaches that tackle the issue of utilizing only an informative subset of channels rather than the complete set of the available sensor channels have been presented recently in the literature. In [9], a sensor channel selection method based on the backward elimination is proposed. The method uses a cost function that is based on the signal to signal-plus-noise ratio with spatial filtering. The evaluation of the selected subset of channels is assessed on three different levels: 1) at a global level, the measure of the signal EEG to noise, 2) at a recognition level, the overall accuracy of the P300 event detection, 3) the accuracy of the speller application. The conducted experiments have shown that selection methods do not consider the spatial filters provide the worst results. In [10], the authors use phase locking values to measure the variability of phase difference between two EEG signals. It basically characterizes the behavioral similarity between two channels. The performance of the selected subset of channels is evaluated by measuring th

This content is AI-processed based on ArXiv data.