Gait pattern classification using compact features extracted from intrinsic mode functions.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Human Motion Analysis

Background:

• Gait patterns are complex, non-linear, and non-stationary signals.
• Empirical Mode Decomposition (EMD) is a suitable method for analyzing such signals.
• Feature extraction is crucial for accurate gait pattern classification.

Purpose of the Study:

• To classify human gait patterns using a novel feature extraction method.
• To evaluate the effectiveness of Discrete Cosine Transform (DCT) on Intrinsic Mode Functions (IMFs) for gait analysis.
• To achieve high classification accuracy for different gait patterns.

Main Methods:

• Human gait data was collected from 52 subjects representing five distinct gait patterns.
• Empirical Mode Decomposition (EMD) was applied to decompose the gait signals into Intrinsic Mode Functions (IMFs).
• Discrete Cosine Transform (DCT) was performed on the IMFs to extract an 8-dimensional feature vector for each gait pattern.

Main Results:

• A compact 8-dimensional feature vector was successfully generated using DCT on IMFs.
• A Gaussian Mixture Model (GMM) was employed for the classification task.
• The proposed method achieved a high overall classification accuracy of 90.2% for the five gait patterns.

Conclusions:

• The combination of EMD and DCT provides an effective feature extraction technique for human gait pattern classification.
• The developed method demonstrates significant potential for accurate and automated gait analysis.
• This approach offers a robust solution for distinguishing between different human gait patterns.