Phase-Dependent EMG Classification for Prosthetic Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current artificial limbs lack the ability to timely and accurately select the correct control mode for joint impedance and motion based on user movement intent, due to limitations in using surface electromyographic (EMG) signals for lower-limb prostheses, which are time-varying and result in low accuracy for pattern recognition.
Innovation Solution
A phase-dependent EMG pattern recognition strategy using Linear Discriminant Analysis (LDA) is implemented to classify user locomotion modes by processing EMG signals from leg muscles, allowing for real-time updates and accurate identification of movement modes for neural-controlled prosthetic legs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If surface EMG signals are used for real-time locomotion mode classification, then the response time is improved, but the classification accuracy deteriorates due to time-varying signal characteristics
Solution Approach 1:
The gait cycle is segmented into distinct phases (stance phase and swing phase) based on footswitch signals. Different classification algorithms are applied to each phase: a fast algorithm using minimal features (RF and VL muscles only) is used for real-time control during swing phase, while a more comprehensive algorithm is used for analysis during stance phase. This segmentation allows the system to achieve real-time response where needed while maintaining accuracy where computationally feasible.
Solution Approach 2:
The invention extracts only the most critical features (RF and VL muscle signals) for real-time classification during swing phase, discarding less critical processing of other muscles (TA, PL, GASL, GASM, SOL) during this phase. This extraction of essential features enables real-time processing while maintaining sufficient classification accuracy for safe prosthetic control.
2Measurement precision
If comprehensive EMG signal processing is used to improve classification accuracy, then the measurement precision is improved, but the computational complexity and processing time increase
Solution Approach 1:
The classification process is segmented into two distinct algorithms: a simple real-time algorithm for swing phase using only RF and VL muscles, and a comprehensive analysis algorithm for stance phase using all available muscle signals. This segmentation reduces computational complexity during the critical real-time swing phase while maintaining accuracy during the less time-critical stance phase.
Solution Approach 2:
During swing phase, the system performs partial processing by considering only the most critical muscle signals (RF and VL) rather than all available EMG channels. This partial action is sufficient for real-time control decisions while significantly reducing computational burden. The full comprehensive processing is reserved for stance phase analysis where more time is available.
3Device complexity
If EMG signals from residual limb only are used, then the device simplicity is maintained, but the neural control information is insufficient for high-level amputations
Solution Approach 1:
The system performs preliminary classification using only residual limb EMG signals (available immediately after amputation) to enable basic real-time control. This preliminary action provides immediate benefit while the patient undergoes TMR surgery and rehabilitation. The system is designed to accommodate future enhancement when additional neural information becomes available through TMR.
Solution Approach 2:
The classification system is designed to be universal, working with whatever neural information is available (whether from residual limb only or from TMR-reinnervated muscles). The same algorithmic framework adapts to different levels of neural input, making the system applicable to patients at any stage of rehabilitation from immediate post-amputation through TMR recovery.
Data Source
AI summary
Apparatus and methods are provided for determining a locomotion mode that can be provided to a controller of a lower prosthesis limb in order to accurately control the prosthesis. One or more prosthesis sensors are provided that break a gait cycle down into a plurality of gait phases. EMG sensors provide signals to a processor that directs them to a gait phase specific classifier that is used to determine a particular locomotion mode for the wearer. With the locomotion mode accurately known, the prosthetic device can be accurately controlled.