Classification of non-weight bearing lower limb movements: Towards a potential treatment for phantom limb pain based on myoelectric pattern recognition

摘要

Research in myoelectric pattern recognition (MPR) for the prediction of motor volition has primarily focused on the upper limbs. Recent studies in the lower limbs have mainly concentrated on prosthetic control, while MPR for lower limb rehabilitation purposes has received little attention. In this work we investigated the viability of a MPR system for the prediction of four degrees of freedom controlled in a near natural or physiologically appropriate fashion. We explored three different electrode configurations for acquiring electromyographic (EMG) signals: two targeted (bipolar and monopolar) and one untargeted (electrodes equally spaced axially). The targeted monopolar configuration yielded overall lower signal-to-noise ratios (SNR) but similar accuracy than those of the targeted bipolar configuration. The targeted bipolar and untargeted monopolar configurations were comparable in terms of SNR and offline accuracy when the same number of channels was used. However, the untargeted configuration tested with twice the channels yielded the best results in terms of accuracy. An advantage of the untargeted configuration is that it offers a simpler and more practical electrode placement. This work is the first step in our long-term goal of implementing a phantom limb pain (PLP) treatment for lower limb amputees based on MPR and augmented/virtual reality.