Identification and classification of upper limb motions using PCA

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• High-dimensional data analysis is crucial in biomedical research.
• Principal Component Analysis (PCA) is a key technique for dimensionality reduction and feature extraction.
• Classifying myoelectric signals is essential for understanding and controlling upper limb movements.

Purpose of the Study:

• To evaluate the utility of Principal Component Analysis (PCA) for classifying upper limb myoelectric signals.
• To compare two distinct input strategies for PCA-based surface electromyogram (SEMG) signal classification.
• To investigate the application of PCA in analyzing SEMG data from upper arm muscles.

Main Methods:

• Acquisition of SEMG data from biceps and triceps brachii muscles in seven subjects.
• Implementation of two PCA-based classification strategies: dual-position myoelectric signal acquisition and sole PCA for SEMG classification.
• Rotation of datasets using class-specific principal component matrices for data decorrelation before feature extraction.

Main Results:

• PCA demonstrated utility in classifying upper limb SEMG signals.
• The study explored two input strategies, highlighting PCA's adaptability in signal processing.
• Decorrelation of measured data via PCA facilitated effective feature extraction for classification.

Conclusions:

• Principal Component Analysis (PCA) is a valuable tool for classifying upper limb myoelectric signals.
• The explored methods provide insights into optimizing SEMG signal analysis for upper limb applications.
• PCA facilitates robust feature extraction essential for accurate myoelectric signal classification.