论文部分内容阅读
在基于表面肌电信号的人机交互系统中,产生的肌肉疲劳降低了系统的稳定性。针对该问题,分析肌肉正常状态和疲劳状态下的肌电信号变化规律,提出一种改进的在线支持向量机增量训练算法。该算法在每次训练SVM(Support Vector Machine)模型时,计算各样本到分类超平面的距离,并以之为条件对不断更新的训练数据进行有条件的选择和遗忘,只留下最大距离1/2以内的数据。通过在线训练不断更新训练样本来获得新的SVM模型,用于适应肌肉疲劳过程中肌电信号的变化,同时防止多次在线训练过程中更新的样本改变训练集间初始边界。最后在智能轮椅上进行验证,实验结果表明:该算法有效减少了肌肉疲劳在人机交互系统中的影响,使得系统能够保持长时间稳定操作。
In the human-machine interaction system based on the surface EMG signal, the resulting muscle fatigue reduces the stability of the system. To solve this problem, we analyzed the regularity of myoelectric changes under normal and fatigue state and proposed an improved incremental training algorithm for online support vector machine. This algorithm calculates the distance between each sample and the classification hyperplane when training the SVM model, and then selects and leaves the continuously updated training data under the condition of only the maximum distance 1 / 2 or less of the data. The new SVM model was obtained by continuously updating the training samples through online training to adapt to the changes of EMG signals during muscle fatigue and to prevent the updated samples during multiple online trainings from changing the initial boundaries between training sets. Finally, it is validated in a smart wheelchair. The experimental results show that this algorithm can effectively reduce the impact of muscle fatigue on the human-computer interaction system and make the system maintain a stable operation for a long time.