Table 1.
Standard EMG onset detection methodologies examined and the iterative settings used for each methodology.
Table 2.
Statistical EMG onset detection methodologies examined and the iterative settings used for each methodology.
Fig 1.
Experimental EMG iterative down-selection process based on root mean square error (RMSE) (Phase 1), clearly aberrant EMG onset detection (Phase 2), and algorithms impacted by the signal-to-noise ratio (Phase 3).
Abbreviations: Raw = raw band-pass filtered EMG; Rect = full-wave rectified EMG; p0 = parameter of the prior on changepoint probability; Prob = posterior probability for EMG onset; LP = low pass filter frequency; SD = standard deviation of time series for EMG onset.
Fig 2.
Simulated EMG iterative down-selection process based on root mean square error (RMSE) (Phase 1) and clearly aberrant EMG onset detection (Phase 2).
Abbreviations: Raw = raw band-pass filtered EMG; Rect = full-wave rectified EMG; p0 = parameter of the prior on changepoint probability; Prob = posterior probability for EMG onset; LP = low pass filter frequency; SD = standard deviation of time series for EMG onset.
Fig 3.
Forest plot of the mean difference between visual EMG onset and algorithm-determined EMG onset for experimentally collected surface EMG.
Circle indicates mean difference and bands are the parametric 95% confidence intervals. The dashed line at ‘0’ corresponds to perfect agreement between methodologies. Intervals crossing ‘0’ indicate no statistical difference between methodologies. Interval width corresponds to the reliability of the estimate. Abbreviations: TKEO = Teager-Kaiser energy operator preconditioning, Low Pass = low pass filter frequency, Thresh = threshold for onset determination, Raw = raw EMG, Rect = full-wave rectified EMG, Seq = Sequential analysis of data, Dist = Distribution, Corr = Correction, p0 = parameter of the prior on changepoint probability.
Fig 4.
Forest plot of the mean difference between known EMG onset and algorithm determined EMG onset for simulated EMG.
Circle indicates mean difference and bands are the parametric 95% confidence intervals. The dashed line at ‘0’ corresponds to perfect agreement between methodologies. Intervals crossing ‘0’ indicate no statistical difference between methodologies. Interval width corresponds to the reliability of the estimate. Abbreviations: Low Pass = zero-lag low pass Butterworth filter, Thresh = threshold for onset determination, Raw = raw EMG, Rect = full-wave rectified EMG, p0 = parameter of the prior on changepoint probability.
Fig 5.
EMG trace in a low noise environment with example onset determinations.
Time series length has been substantially cropped, focusing on the time of onset, in order to increase the visibility of onset determination for various methods. Abbreviations: Rect = full-wave rectified EMG; p0 = parameter of the prior on changepoint probability; LP = low pass filter frequency; SD = standard deviation of time series for EMG onset; Thresh = threshold for EMG onset.
Fig 6.
EMG trace in a moderate noise environment with example onset determinations.
Time series length has been substantially cropped, focusing on the time of onset, in order to increase the visibility of onset determination for various methods. Abbreviations: Rect = full-wave rectified EMG; p0 = parameter of the prior on changepoint probability; LP = low pass filter frequency; SD = standard deviation of time series for EMG onset; Thresh = threshold for EMG onset.