图共 8个 表共 3
    • 图  1  串并行卷积神经网络结构图

      Figure 1. 

    • 图  2  Emotiv脑电采集仪

      Figure 2. 

    • 图  3  Emotiv脑电采集仪电极安放位置

      Figure 3. 

    • 图  4  单次脑电信号采集过程

      Figure 4. 

    • 图  5  设定不同的阈值$\alpha $时识别率情况

      Figure 5. 

    • 图  6  设定不同的阈值$\beta $时识别率情况

      Figure 6. 

    • 图  7  采用不同处理方法的识别准确率对比

      Figure 7. 

    • 图  8  智能轮椅系统结构图

      Figure 8. 

    • 算法CSPACSPDBNCNNSTFT-CNNSPCNN本文CEMD-SPCNN
      S0165.0077.5087.0886.2588.7590.4293.33
      S0281.6782.9287.5087.9289.1791.2594.17
      S0398.3397.0895.8395.8396.6797.0899.16
      S0476.2578.3383.3385.4285.4286.2589.58
      S0595.4296.2593.7591.6792.5094.1796.67
      均值83.3386.4189.5089.4290.5091.8394.58
      方差190.0191.8826.5118.6118.1816.6213.02

      表 1  不同算法对5受试者脑电信号的识别准确率(%)

    • 算法ChinGanCoyleCSPACSPDBNCNNSTFT-CNNSPCNN本文CEMD-SPCNN
      B0170.0071.0060.0066.5667.5066.5672.2275.0076.3980.56
      B0261.0061.0056.0057.8155.3162.5061.0361.7663.2464.71
      B0361.0057.0056.0061.2562.1960.0061.1162.5062.5064.58
      B0498.0097.0089.0094.0694.6996.8798.6598.6599.3299.32
      B0593.0086.0079.0080.6376.8882.1986.4887.1687.8488.51
      B0681.0081.0075.0075.0075.9477.5079.1780.5681.2583.33
      B0778.0081.0069.0072.5071.2576.5678.4777.0879.1781.25
      B0893.0092.0093.0089.3889.3888.7586.1886.1886.8490.13
      B0987.0089.0073.0085.6381.2585.9481.2582.6484.0386.81
      均值80.2279.4472.2275.8674.9377.4378.2879.0680.0682.13
      方差192.19190.03181.69158.75157.50155.85147.92138.63137.52129.78

      表 2  不同算法对BCI competition IV 2b数据集的识别准确率(%)

    • 操作直行左转右转
      S01968486
      S02949082
      S03988892

      表 3  各类操作在线识别准确率(%)