Motor imagery classification method combining ensemble learning and independent part analysis

• Main Authors: LI SHISHEN, ZHOU BANGYAN, WU XIAOPEI, LYU ZHAO, ZHANG LEI, ZHANG YUJUN, ZHANG CHAO, DU CHENXIAO, CHEN WENWEI, GUO XIAOJING
• Format: Patent
• Language: Chinese; English
• Published: 01.12.2020
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a motor imagery classification method combining ensemble learning and independent part analysis. Based on different single test data, an ICA spatial domain filter is constructed to perform spatial domain filtering and feature extraction on the data, different base classifiers are constructed by using simple classification rules, and then a new motor imagery electroencephalogram signal classifier is constructed by using a voting method of ensemble learning in combination with a strategy. The integrated classifier designed by the invention has a relatively stable recognition rate, multiple base classifiers are integrated for learning classification, generalization performance superior to that of a single classifier can be obtained, and the classification recognition rate of EEG signals is improved. In addition, in migration tests of data collected among different subjects and in different periods of the same subject, stable performance is also embodied, and betterpracticability is achiev