The invention relates to the technical field of intracranial electroencephalogram
signal processing, in particular to an intracranial electroencephalogram multivariable decoding method and
system based on sparse
logistic regression, and the method comprises the steps: obtaining an intracranial electroencephalogram original
signal collected by an implantable
electrode, and sequentially carrying out the filtering,
power frequency interference removal, bad contact removal and re-reference preprocessing of the intracranial electroencephalogram original
signal; calculating energy measurement of a whole brain
electrode through time-
frequency analysis, and forming three-dimensional feature tensors of a space domain M, a
time domain T and a
frequency domain F; expanding the three-dimensional feature
tensor into a
feature matrix based on a
multivariate analysis method, and selecting an acceleration path of matrix operation according to a ratio of a dimension D to a
sample number N; and performing feature subset selection based on sparse
logistic regression, inputting the selected feature subset into a classifier, and generating a neural decoding instruction, thereby solving the problems of large data volume, calculation
delay, dimension disaster and low scene generalization in the prior art.