Firpi H Alexer, Vogelstein R Jacob
Johns Hopkins University/Applied Physics Laboratory, Laurel, MD 20723-6009, USA.
Annu Int Conf IEEE Eng Med Biol Soc. 2011;2011:6556-9. doi: 10.1109/IEMBS.2011.6091617.
This manuscript proposes a particle swarm-based feature extraction to monitors brain activity with the goal of identifying correlate cognitive states and intensity of a task. This in turn would allow us to develop a pattern recognition system that will classify such cognitive states and thus to redistribute the workload to other subjects. In this abstract, we present a recognition system that employ multiple features from different domains, a feature selection method using a Particle Swarm Optimization (PSO) search algorithm while the classification is provided using a k-nearest neighbor. Through this approach, we are able to achieve an averaged classification accuracy of 90.25% on held-out, cross-validated data among the eight subjects.
本手稿提出了一种基于粒子群的特征提取方法，用于监测大脑活动，目的是识别与任务相关的认知状态和强度。这反过来将使我们能够开发一个模式识别系统，该系统将对这些认知状态进行分类，从而将工作量重新分配给其他受试者。在本摘要中，我们展示了一个识别系统，该系统采用来自不同领域的多个特征，一种使用粒子群优化（PSO）搜索算法的特征选择方法，同时使用k近邻进行分类。通过这种方法，我们能够在八名受试者留出的交叉验证数据上实现平均90.25%的分类准确率。
