Abstract—We present a wavelet two-directional twodimensional principal component analysis (WT2D2PCA) method for the efficient and effective extraction of essential feature information from high-dimensional signals. Wavelet multi-scale matrices constructed in the first step incorporate the spatial correlation of sub-band signals among channels. In the second step, the two-directional two-dimensional principal component analysis operates on the multi-scale matrices to reduce the dimension, rather than vectors in conventional principal component analysis (PCA). Results are presented from an experiment to classify 20 hand movements using 89-channel myoelectric signals (MES) recorded in stroke survivors, which illustrates the efficiency and effectiveness of the proposed method for high-dimensional signal pattern recognition.
Index Terms—Time-frequency analysis, wavelet transform, two-directional two-dimensional principal component analysis, myoelectric signals, pattern classification.
H. B. Xie is with ARC Centre of Excellence for Mathematical and Statistical Frontiers, Queensland University of Technology, Brisbane, QLD 4000, and Jiangsu Provincial Key Laboratory for Interventional Medical Devices, Huaiyin Institute of Technology, Huaian, China.
Jianhua Wu and Lei Liu are with Jiangsu Provincial Key Laboratory for Interventional Medical Devices, Huaiyin Institute of Technology, Huaian, China (email: hongbo.xie@qut.edu.au).
Cite: Hongbo Xie, Jianhua Wu, and Lei Liu, "Pattern Classification of High-Dimensional Myoelectric Signals Using Wavelet Two-Directional Two-Dimensional Principal Component Analysis," International Journal of Machine Learning and Computing vol.6, no. 1, pp. 67-70, 2016.