基于SSVEP的空间机械臂脑机交互方法

A brain computer interaction method of space manipulators based on SSVEP

摘要:

为实现航天员与空间机械臂的脑机交互,针对稳态视觉诱发电位脑机接口(SSVEP-BCI),提出一种基于卷积神经网络的SSVEP信号分类方法。该方法以SSVEP信号经过快速傅立叶变换的特征为输入,经过三层卷积层、全连接等操作实现信号的分类识别。采用清华大学Benchmark数据集对该方法进行测试,在1秒的时间窗口下,平均分类准确率为99.07%,平均信息传输率为149.24b/min,均明显高于采用典型相关分析或滤波器组典型相关分析的方法。实验对比分析表明,该方法针对短时间窗口的SSVEP信号具有较好的目标分类效果。最后,使用分类后的信号作为控制信号,对仿真环境下的空间机械臂进行操作,实现人和空间机械臂的脑机交互。

Abstract:

In order to realize the brain computer interaction between astronauts and space manipulators, aiming at the steady-state visual evoked potential brain-computer interface (SSVEP-BCI), the SSVEP signals classification method based on convolutional neural network is proposed. This method takes the characteristics of SSVEP signals after fast Fourier transform as the input, and realizes the classification and recognition of SSVEP signals through three-layer convolution and full connection. The benchmark data set of Tsinghua University is used to verify the method. Under the time window of 1 second, the average classification accuracy is 99.07% and the average information transmission rate is 149.24b/min, which are significantly higher than those using canonical correlation analysis or filter bank canonical correlation analysis. The comparison result shows that the proposed method has good target classification effect for short-time SSVEP signals. Finally, the recognized signals are used as the control signals to operate the space manipulator in the simulation environment to realize the brain computer interaction between human and space manipulators.