User intent prediction with a scaled conjugate gradient trained artificial neural network for lower limb amputees...

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Artificial Intelligence in Healthcare

Background:

• Powered lower limb prostheses enhance mobility for amputees, often featuring pre-programmed modes for activities like stair climbing.
• Seamless mode transitions in prostheses are crucial for user function, with pattern classification showing high accuracy.
• Current methods require time-consuming subject-specific training data, limiting practical application.

Purpose of the Study:

• To investigate the efficacy of an artificial neural network (ANN) for classifying gait intention in lower limb amputees.
• To evaluate the reduction in training time using subject-independent datasets compared to subject-dependent datasets.
• To compare the performance of an ANN against a linear discriminant analysis (LDA) classifier for gait intention recognition.

Main Methods:

• An intent recognition system was developed using an artificial neural network (ANN) with a scaled conjugate gradient learning algorithm.
• Gait intention was classified using both subject-dependent and subject-independent datasets from six unilateral lower limb amputees.
• The ANN's performance was compared against a linear discriminant analysis (LDA) classifier.

Main Results:

• The ANN demonstrated significantly lower classification error (P<0.05) than LDA for all user-dependent step types.
• The ANN also outperformed LDA in classifying transitional steps using user-independent datasets.
• Both LDA and ANN classifiers achieved fast decision-making times (1.29 ms for LDA, 2.83 ms for ANN).

Conclusions:

• Artificial neural networks (ANNs) offer a suitable and accurate method for offline classification in prosthesis gait prediction.
• The use of subject-independent datasets with ANNs shows promise in reducing prosthesis training time.
• ANNs provide a robust approach for enhancing the functionality and user experience of powered lower limb prostheses.