What do the figures mean?

They illustrate various aspects of EMG decomposition. They are briefly described here.

What if my signals are already high-pass filtered?

If your signals are already high-pass filtered, then you can decompose them without applying any additional high-pass filtering (i.e., the high-pass popup left at "unfiltered"). In this case, though, you should change the "MUAP width" setting in the Preferences menu. This setting specifies the length of the templates that EMGlab uses when working with the unfiltered signal. The default value of 60 ms is appropriate for MUAPs with low-frequency components, but it is makes template matching and automatic decomposition slow and inefficient. If your MUAPs are already spiky, then reduce the "MUAP width" setting to a more appropriate value, such as 5 or 10 ms. Note that this setting does not take place immediately. You will have to re-open your data file in order for the change to take effect.

Why doesn't the automatic decomposition algorithm do a good job on my signal?

If you are using the automatic decomposition algorithm with the high-pass filter popup set to "unfiltered," then make sure that you have set the "MUAP width" setting in the Preferences menu appropriately (see "What if my signals are already high-pass filtered?"). Sometimes the algorithm may have difficulty identifying templates in the first place, but it may do a better job of recognizing them once you have established them by hand. Signals can vary considerably not only in terms of complexity, but also in terms of the characteristics of the spikes and the baseline noise. The algorithm tries to pick thresholds and tolerances for template matching based on the characteristics of each particular signal. However, it might not always do this in an appropriate way. Greater sharing of data files between labs will help us develop algorithms that are more robust across a wider range of signals.

How can I set the detection threshold to prevent the automatic decomposition algorithm from trying to identify tiny spikes?

You can't in the current version. The algorithm picks the detection threshold automatically, based on an estimate of the baseline noise. Sometimes this threshold is very small, and the program tries to identify very tiny spikes. In some signals you might be able to fully decompose these tiny spikes, but in other signals it's not worth the effort. More user control of this threshold will be included in future versions of EMGlab. For now, you will just have to delete any small units that you don't want to decompose.

Why are some signals so much longer than others?

Different investigators use different sampling rates, ranging from 10 kHz to 50 kHz. In theory, a rate around 10 kHz is adequate to capture all the information in the signal, but some decomposition software require that signals be oversampled in order to achieve adequate temporal precision. EMGlab can work with signals whether they are oversampled or not because it uses interpolation in its template-matching algorithms. In fact, EMGlab down-samples overrsampled signals when it loads them to make them more manageable. (You can control the amount of down-sampling by means of the "Nyquist Rate" setting in the Preferences Menu.) Down-sampled versions of many signals are provided in the database. These files are much shorter than the original ones, but they contain all the same information. They can be used with EMGlab, but not with the decomposition software for which the signal was originally intended.

Why do I sometimes get flat templates when I load an annotation file?

EMGlab computes the templates by averaging the occurrences of each unit that occur within the interval shown in the firing panel. If there are less than three occurrences of the unit in that interval, then the template is set to zero. To get the correct templates, zoom the firing panel out to include a suitable number of occurrences of each unit.