Motion Intent Recognition in Intelligent Lower Limb Prosthesis Using One-Dimensional Dual-Tree Complex Wavelet...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Human-Computer Interaction

Background:

• Motion intent recognition for lower limb prostheses is vital for user mobility.
• Traditional methods struggle with unstable statistical features during gait transitions.
• Gait instantaneous conversion between steady states presents a significant challenge.

Purpose of the Study:

• To introduce a novel method for motion intent recognition in lower limb prostheses.
• To address the limitations of traditional statistical feature-based approaches.
• To improve the accuracy and stability of recognizing gait transitional patterns.

Main Methods:

• Application of the one-dimensional dual-tree complex wavelet transform (1D-DTCWT).
• Utilizing the local analysis ability of wavelet transform to amplify gait variation characteristics.
• Leveraging translation invariance and direction selectivity of 1D-DTCWT for continuous feature extraction.

Main Results:

• High recognition accuracy achieved: 98.91% (steady), 98.92% (transitional), 97.27% (total) for able-bodied subjects.
• Excellent accuracy for an amputee subject: 100% (steady), 91.16% (transitional), 90.27% (total).
• Demonstrated superior performance in exploring gait instantaneous conversion compared to state-of-the-art methods.

Conclusions:

• The proposed 1D-DTCWT method offers a more stable and accurate approach to motion intent recognition.
• This technique effectively captures the continuous features of human lower limb movements.
• The findings indicate a significant advancement in lower limb prosthesis control and user experience.