*2.3. Data Analysis*

#### 2.3.1. Data Preprocessing

Surface EMG and force data were processed offline in MATLAB (MathWorks, Natick, MA, USA). A 6th order Butterworth (10–500 Hz) was applied to the EMG signals. The power line interference in the EMG signal was eliminated using a spectrum interpolation algorithm [34]. Force signals were manually inspected to select a relatively stable 5 s segment. Surface EMG and force signals within the epochs were extracted for further analysis.

#### 2.3.2. Calculation of MFCV

Prior to the analysis of MFCV, EMG signals were differentiated between consecutive channels to generate 19 channels of bipolar signals. The IZ was determined by either visual inspection or analysis of the bipolar signals. The IZ was estimated to be the channel with the lowest amplitude, or between the channels that demonstrated reverse signal polarity and a clear pattern of bidirectional signal propagation from the IZ channel to the tendons [35]. The MFCV was determined based on detection of the temporal delay between adjacent single differential channels. Specifically, MFCV was defined and computed as *d*/*τ*, where *d* is the inter-electrode distance between the channels and *τ* is the time delay between two channels (calculated from cross-correlation analysis). Channels containing IZ or adjacent to IZ were excluded for MFCV estimation. The MFCV calculated at each contraction level was averaged over all contraction levels for further comparison between groups.

#### 2.3.3. EMG–Force Relation

The force signal from individual trials was averaged over the selected epoch. The corresponding surface EMG amplitude was obtained by calculation of the root mean square (RMS) value from each channel. Channels close to the proximal and distal tendons were excluded from analysis. The channel producing the maximal RMS values (by evaluating the eight target force contractions and the MVC) was used for estimation of the EMG–force relation. Next, the RMS values from the selected channel were averaged across the 2 trials for each force level. The force was also averaged across the trials. The RMS and force values were then normalized to the MVC values for determination of the EMG–force relation.

The EMG–force relation was estimated in each SCI and control subject. Given that curvilinear EMG–force relation has been widely reported for large muscles such as BB [36,37], we applied quadratic fitting to describe the BB EMG–force relation. The quadratic equation was expressed as: *y* = a*x* <sup>2</sup> + b*x* + c, where *x* represents force and *y* represents EMG amplitude. The coefficient of determination (R<sup>2</sup> ) for quadratic fitting was calculated for each subject.
