Upper Arm Motion High-Density sEMG Recognition Optimization Based on Spatial and Time-Frequency Domain Features

Table 2

Spatial feature extraction based on multiclass CSP.

---

Input: covariance matrices

---

(1) Perform joint approximate diagonalization

(2) For each column , of scale , estimate mutual information according to

(3) Choose the columns of with highest mutual information

(4) Acquire the raw channel number of the columns of W with highest mutual information

---

Output: preprocessing matrix and channel number with highest mutual information

---