Synergy Extrapolation

Intro

Synergy Extrapolation (SynX) is a tool to fill in the gaps for incomplete EMG data sets by predicting the EMG signals for muscles that do not have measured data. With current technologies, it is uncommon to have a complete data set with every muscle measured individually due to limitations in EMG placement and the number of channels that can be measured simultaneously. SynX compensates for this by taking in the available data and decomposing it into a small number of time varying signals and assigning corresponding signal weights to each muscle. It is then able to use this framework to fill in the missing EMG signals for the model.

How it Works

SynX breaks down a large number of time varying EMG signals into a small number of time varying “synergy” signals along with a set of constant weights for each muscle to describe how that synergy affects the individual muscle excitations. It can do this using either a principle component analysis (PCA) approach or Non-Negative Matrix Factorization (NMF) in order to minimize errors between the model’s motion and the original data set. The user can select which method they would prefer in the MTP xml file. Using these results, the SynX tool steps through the model’s motion to determine both the unmeasured muscle excitations and residual excitations for the measured muscles.

Importance

This new method of predicting synergy values allows for significantly greater accuracy. Previous methods such as static optimization relied on making many assumptions about the subject, which would not always be accurate. For example, one common assumption was that the model will always optimize its muscle activations to minimize the energy cost. However, this neglects many situations in which opposing muscles are activated simultaneously during a motion for stabilization. SynX is able to account for this because it doesn’t rely on making assumptions about how the body is optimizing its muscle activations. In fact, in a paper by Bianco et al. they found that using just 8 synergy groups could accurately predict over 90% of the muscle variance throughout the body.