Common Spatio-Time-Frequency Patterns for Motor Imagery-Based Brain Machine Interfaces

Algorithm 1

Design of the FIR filters, the spatial weights, and the time windows.

Input:  𝒞₁ and 𝒞₂: the sets of observed signals.

Parameters:  𝐹:  the number of FIR filters, 𝑃: the filter order,

ℬ: the set of the candidates for the time window.

Output:  {𝐰_𝑖}^𝐹_𝑖=1: the spatial weights, {𝐡_𝑖}^𝐹_𝑖=1: the vectors of

FIR filter coefficients, {𝐛_𝑖}^𝐹_𝑖=1: the time window.

for  𝑖=1,…,𝐹  do

 Initialize 𝐡_𝑖 and 𝐛_𝑖.

 Set the index of iteration as 𝑘=0.

 repeat

  𝑘←𝑘+1.

  Update 𝐰_𝑖 by solving (11).

  Update 𝐡_𝑖 by solving (13).

  Update 𝐛_𝑖 by solving (19).

  Calculate cost, 𝐶_𝑘 from the cost function, 𝐽(̂𝒫_𝑖).

 until 𝐶_𝑘−𝐶_𝑘−1 is sufficiently small.

end for