基于P300的字符输入脑机接口研究

发布时间：2018-07-11 09:47

本文选题：脑机接口 + P300　；参考：《华南理工大学》2015年硕士论文

【摘要】：脑机接口(Brain-computer interface,BCI)是指在不依赖于组织肌肉的情况下,直接将采集的脑信号转化为相应的机器命令去控制外部设备,从而完成人机交互的过程。随着信息时代的快速发展,BCI技术也日益成为备受关注的课题。它在大脑与外部环境之间建立了一个独立的交流通道,为运动障碍的残疾人提供了一种与外界交流的方式。此外,BCI的研究也应用于多个领域,如医学、控制、人工智能和脑科学等,而本文研究的基于P300的BCI系统也是脑机接口最为常见的一个应用。然而,目前基于P300的BCI系统虽然有不少的研究成果,但在实际应用中仍然存在系统的信息传输速率较低和如何检测异步系统的空闲状态等问题。针对上述问题,本文主要做了以下两个方面的研究工作:首先,本文提出了一种缩短刺激序列的动态停止准则(dynamic stopping criterion,DSC)P300检测方法,该算法在基于Bayesian框架下计算每个字符成为目标字符的后验概率,并将此概率值与设定的阈值相比。若大于阈值则闪烁序列停止,直接输出该目标字符;否则将当前刺激序列中所有字符的后验概率值进行递减排序,在刺激序列中删除概率值较小的字符,从而缩短了闪烁字符序列,减弱非目标字符闪烁对受试者的干扰。由实验结果表明,该方法在保证较高准确率的前提下,明显提高了字符的输入速度。其次,本文初步探讨了基于异步P300的字符输入BCI系统,该系统根据是否能检测到P300信号判断用户是否正在有输入,与传统的同步P300 BCI系统相比,增加了空闲状态。而对于控制状态和空闲状态的检测,本文介绍了两种方法,一种是基于统计学分析的异步模式检测方法,该方法是基于Bayesian框架下计算闪烁序列在当前状态下的后验概率。若此概率值大于设定的阈值,则判断为控制状态并根据DSC算法输出目标字符;另一种方法是基于组合分类器的异步模式检测,该方法是采用两级分类器建立分类器模型,再根据第二级分类器预测的label值来判断。若label值为1则判断为控制状态,然后同样根据DSC算法输出目标字符。通过实验数据分析,结果表明异步模式的两种状态在系统中能90%被准确的检测出来。
[Abstract]:Brain-Computer Interface (BCI) refers to the process of human-computer interaction by converting the collected brain signals directly into corresponding machine commands to control the external equipment without relying on the tissues and muscles. With the rapid development of information age, BCI technology has become a subject of great concern. It establishes an independent communication channel between the brain and the external environment, and provides a way to communicate with the outside world for the disabled with dyskinesia. In addition, the research of BCI is also applied in many fields, such as medicine, control, artificial intelligence and brain science. The BCI system based on P300 is also the most common application of BCI. However, although there are many research achievements in P300-based BCI system, there are still some problems in practical application, such as low information transmission rate and how to detect the idle state of asynchronous system. In order to solve the above problems, this paper mainly studies the following two aspects: firstly, this paper presents a dynamic stop criterion (dynamic stopping criteria P300 detection method for shortening stimulus sequences. The algorithm calculates the posteriori probability that each character becomes the target character based on Bayesian framework, and compares the probability value with the set threshold. If the threshold value is greater than the threshold value, the flicker sequence stops, and the target character is output directly. Otherwise, the posteriori probability of all characters in the current stimulus sequence is decremented and the characters with lower probability are deleted in the stimulus sequence. Thus, the flicker character sequence is shortened and the interference of non-target character flicker is reduced. The experimental results show that the proposed method can improve the input speed of characters on the premise of high accuracy. Secondly, this paper preliminarily discusses the character input BCI system based on asynchronous P300, which can detect P300 signal to determine whether the user is having input or not. Compared with the traditional synchronous P300 BCI system, the system increases the idle state. For the detection of control state and idle state, this paper introduces two methods, one is asynchronous pattern detection method based on statistical analysis, and the other is based on Bayesian framework to calculate the posteriori probability of flicker sequence in the current state. If the probability value is greater than the set threshold, the control state is judged and the target character is outputted according to DSC algorithm. Another method is asynchronous pattern detection based on combinatorial classifier, which uses two-level classifier to build classifier model. Then according to the label predicted by the second level classifier to judge. If the label value is 1, the control state is determined, and the target character is also output according to the label algorithm. Through the analysis of experimental data, the results show that the two states of asynchronous mode can be accurately detected by 90% in the system.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R319;TN911.7

【相似文献】