面向假肢的场景动画稳态视觉诱发脑控方法

Brain-Controlled Prosthesis Manipulation Based on Scene Graph-SSVEP

针对现有脑控假肢技术的控制精度低、稳定性差的问题,基于稳态视觉诱发电位(SSVEP)的产生机理,提出一种由场景动画稳态视觉诱发的脑控新范式。该范式以正常人或残疾人的生活场景为刺激源蓝本,根据智能假肢的控制目标,在将生活场景分解为对应的独立刺激场景图、且对其进行灰度标准化处理后,采用方波调制模式对一组对比鲜明的黑白反转色图片进行视觉刺激,由此诱发出一种基于场景动画的SSVEP;进而,通过对场景动画的SSVEP神经传导过程进行数学建模与仿真,建立了一种基于典型相关分析(CCA)的脑电信号处理方法。在专用于场景动画SSVEP的智能假肢脑控平台上进行实验,系统的平均正确率为91.41%,平均信息传输率为15.32bit/min,其最高平均识别率达到了98.44%。实验结果表明:该方法可将正常人生活场景图与传统稳态视觉诱发方法进行结合,不仅能够提高假肢动作的平均识别精度和信息传输率,而且具备可降低使用者视觉疲劳的作用。

To solve the problems of low recognition rate and stability of traditional steady-state visual evoked potential（SSVEP）methods,a new scene-graph SSVEP approach was proposed.The new scene-graph SSVEP paradigm uses life scenes of normal or disabled persons as the stimulus source.According to the target of prosthesis control,the life scenes are decomposed into corresponding stimulation scene graphs through standardization process of gray scales.After that,a set of contrasting black-and-white reversal scene graphs are obtained for visual stimulation.The new scene-graph SSVEP is evoked by a square pulse modulation with different frequencies and different scene images.Furthermore,the mathematic model of the scene-graph SSVEP nerve conduction is simulated.For recognizing various EEG signals from different scene graph stimulations,canonical correlation analysis（CCA）is used to turn EEG features into control commands.Finally,a brain-controlled prosthesis manipulation platform was built and the new strategy was verified by experiments.Results show that the information transfer rate is approximately 15.34bit/min,the average accuracy is 91.4% and the highest accuracy is up to98.44%.It is demonstrated that this method combining normal life scene graphs with traditionalSSVEP methods can improved the average recognition accuracy and information transfer rate of prosthesis,and reduce user＇s visual fatigue.