Short latency hand movement classification based on surface EMG spectrogram with PCA.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Surface electromyography (sEMG) is crucial for developing advanced motor prostheses.
• Effective hand gesture recognition relies heavily on the quality of extracted EMG features.
• Current methods face challenges in accuracy and robustness, especially during continuous gesture classification.

Purpose of the Study:

• To evaluate EMG spectrograms as a superior feature for discriminating hand gestures from sEMG data.
• To assess the impact of principal component analysis (PCA) on EMG spectrogram features for dimensionality reduction.
• To improve classification accuracy and reduce error rates in continuous hand gesture recognition.

Main Methods:

• Utilized the Ninapro database containing 12-channel sEMG data from 40 subjects performing 50 hand movements.
• Extracted EMG spectrogram features from sEMG signals.
• Applied principal component analysis (PCA) for dimensionality reduction of EMG spectrogram features.
• Compared classification performance against traditional time-domain features.

Main Results:

• EMG spectrograms, after PCA, significantly improved hand gesture classification accuracy by approximately 10% compared to time-domain features.
• The proposed method demonstrated enhanced performance for recognizing 50 distinct hand movements, including subtle finger motions and varying force levels.
• An approximate 12% reduction in error rate was observed during gesture transition phases, indicating improved robustness.

Conclusions:

• EMG spectrograms represent a highly effective feature set for multi-class hand gesture recognition from sEMG.
• Dimensionality reduction using PCA on EMG spectrograms enhances classification performance and efficiency.
• This approach offers a promising pathway for more robust and accurate continuous gesture recognition in motor prosthesis applications.