Removal of Muscle Artifacts from Single-Channel EEG Based on Ensemble Empirical Mode Decomposition and Multiset Canonical Correlation Analysis

Algorithm 1

The EEMD-MCCA Algorithm.

Input: the single-channel EEG signal 𝐱 with size 1×𝑇.

Output: the reconstructed EEG signal 𝐱 after muscle artifact removal.

The First Step:

(1) for  𝑖=1∶𝐼  do

(2)  Add independent identically distributed white noise to the single-channel EEG 𝐱;

(3)  Apply EMD to the above noisy signal and derive a set of IMFs by (1)–(10), denoted as 𝐅_𝑖;

(4) end  for

(5) Obtain an ensemble of IMF sets 𝐅_𝑖’s;

(6) Calculate a set of averaged IMFs as the final decomposition, that is 𝐗=∑^𝐼_𝑖=1𝐹_𝑖/𝐼;

The Second Step:

(7)   𝑀 temporally delayed versions of the matrix 𝐗 are generated according to (12), that is 𝐗_𝑚,𝑚=1,2,…,𝑀;

(8)   Apply MCCA to the 𝑀 data sets and extract the underlying sources 𝐒 in 𝐗;

(9)   Set the sources 𝐒 corresponding to muscle artifacts (with low autocorrelation) to zero;

(10)   Return the cleaned multichannel signals 𝐗 by passing the source matrix through the mixing matrix 𝐀;

(11)    Reconstruct the single-channel EEG signal 𝐱 by summing the recovered IMFs in the matrix 𝐗.