Multi-modal signal-based motor imagery intention recognition method

• Main Authors: XIE PING, LI ZENGYONG, XU XIANGYUAN, FENG LUFENG, WANG KUNYU, HE QUN, CHEN XIAOLING, JIANG GUOQIAN
• Format: Patent
• Language: Chinese; English
• Published: 21.01.2022
• Subjects: DIAGNOSIS; HUMAN NECESSITIES; HYGIENE; IDENTIFICATION; MEDICAL OR VETERINARY SCIENCE; SURGERY

The invention discloses a multi-modal signal-based motor imagery intention recognition method. The method comprises the steps: designing an asymmetric multi-task convolutional network through the temporal-spatial resolution complementarity of an electroencephalogram signal and a near-infrared signal, and the correlation of multiple tasks, extracting the space-time features of the two signals in a parallel mode, and performing feature layer fusion on the two signals in a cascading mode and then sending to an auxiliary measurement task and a main classification task. The technical means fuses a multi-task learning method, therefore, more representative features can be extracted, and compared with a traditional convolutional network model and variants thereof, the classification performance can be enhanced, the model generalization ability is improved, and a new way is provided for motor imagery classification and recognition. 本发明公开了一种基于多模态信号的运动想象意图识别方法，利用了脑电信号与近红外信号的时空分辨率互补性，以及多任务的相关性，设计非对称的多任务卷积网络以并行方式提取两种信号的空