SEMG applications use features extracted from surface electromyography (SEMG) data as inputs. The features often exhibit high variability across subjects, activities, and even across segments from a single contraction. The purpose of this work was to provide insights into SEMG feature variability to enable better data collection and interpretation. Variability in mean frequency (MF) and mean absolute value (MAV) was measured in simulation across stationary segments to estimate baseline variability. Then variability in the features was measured during two in vivo protocols involving static contractions: a single contraction ‘within-trial’ test and a set of contractions yielding a ‘between-trial’ test. For the ‘within-trial’ test, variability in MAV increased significantly from baseline, while the variability of MF did not (ANOVA, α=0.05, p<0.001, p=0.35, respectively). For the ‘between-trial’ test, variability in MAV increased further, and MF significantly increased as well (ANOVA, α=0.05, p<0.05, p<0.01). A sensitivity analysis conducted in simulation was used to infer sources for variability beyond the baseline, and the number of active motor units emerged as the likely source for ‘within-trial’ increases, while conduction velocity and fibre depth emerged as likely contributors to ‘between-trial’ increases, along with the number of active motor units. Feature variability can also be affected by electrode position during measurement. Focusing on spectral energy, bandwidth characteristics, and signal amplitude characteristics, five features were examined for use as an index to avoid the effects of innervation zone and tendon regions. Band power ratio (BRP), MF and absolute area of a normalized action potential (AANAP) demonstrated promising results by correctly identifying >80% of poorly positioned electrode channels.
Zircon - This is a contributing Drupal Theme Design by WeebPal.
