A novel hidden Markov model-based pattern discrimination method with the anomaly detection for EMG signals

Area of Science:

• * Biomedical Engineering
• * Machine Learning
• * Signal Processing

Background:

• * Accurate classification of sequential patterns is crucial in various applications, including biomedical signal analysis.
• * Existing methods often struggle with classifying patterns from unlearned or undefined classes.

Purpose of the Study:

• * To propose a novel sequential pattern recognition method for classifying both learned and unlearned classes.
• * To incorporate probability density functions of unlearned classes into a hidden Markov model for enhanced classification.
• * To validate the method's effectiveness in motion classification using electromyogram (EMG) signals.

Main Methods:

• * Development of a sequential pattern recognition technique utilizing hidden Markov models.
• * Integration of probability density functions for unlearned classes into the model parameter estimation.
• * Experimental validation using EMG signals from three subjects performing eight forearm motions.

Main Results:

• * Achieved high classification performance for learned motions (90.13%) and unlearned motions (91.25%).
• * Demonstrated superior performance compared to previous pattern recognition approaches.
• * Successfully classified undefined classes through model parameter estimation.

Conclusions:

• * The proposed method effectively enhances sequential pattern recognition for both learned and unlearned classes.
• * The technique shows significant potential for applications in motion classification and disease diagnosis support.
• * The incorporation of unlearned class information improves overall classification accuracy.