Gait data: step count and average waveform

[MJG3]

I am currently evaluating DIAdem to determine if it will be usable for my task. I have human gait force/moment data to analyze. In particular, I am analyzing force data similar to force plate data. Data is anticipated to be collected over periods of days so they will be relatively large files. First, I need to segment the data according to activity (inactive, standing, walking, localized movement). I have figured out how to do this manually using flags and visual inspection. If I go forward with the software, I will want to automate this process but my current questions are regarding the analysis of the segmented data.

(1)    For the walking and localized movement segments, I want to know the number (count) of gait steps. One complete step is illustrated in the pdf between the two black solid lines. The gait movement typically has a double peak appearance but that is not always the case, sometimes it resembles a single peak. The peak search does not necessary give me what I want.

(2)    For the walking segment, I want an average gait step waveform. Take all the individual steps and display an average waveform (one complete step).

(3)    For each activity segment, I want to know the total duration for the activity. The sampling rate is not necessary constant and might fluctuate or change. Since the segmented time data references the relative time location, I haven’t been able to figure out how to represent the total time especially since the sample rate changes.

I have attached a sample of the data which includes the complete force data and manually segmented activity data. I am not experienced with VBscript writing. Suggestions/assistance with any of these tasks will be helpful or indication if any of these tasks are not feasible with DIAdem.

Thanks

[Trey_C]

Hi MJG3,

What you are trying to do is definitely possible with DIAdem and while it is a lot easier than it would be in other languages, it isn't necessarily simple - especially if you don't have much experience with VBscripting.  I took a look at the data that you attached and you will definitely need to create an algorithm to filter out each different section of your data to create separate graphs.  In order to actually create these graphs, it is probably easier to record a macro in the Scripting tab that creates a subset of data and then adds it to your report (or wherever you need it) than to write it from scratch.  Then, from there, I would use the algorithm you develop to determine where to put each data point.

As far as the actual algorithm goes, I have a general idea of how you might want to do this.  You could use a 'filter' (if/else if) statements that look at the following parameters to determine whether you are walking, standing, localized, or inactive.

If you are walking, it looks like you could use the frequency and amplitude of your signal to determine if you are walking (If Frequency > limit && Amp > limit2)

Next, I would check to see if you are standing by determining if the change in amplitude over a certain period of time is less than a certain amount but the amplitude is greater than a set limit. I might suggest using an offset of your localized data followed by an integral to get the area under the curve to determine the change.  

To determine if you are 'inactive', I would use the same parameters as standing but you would want to look below the set limit.

Finally, any data that is not accounted for is 'localized' which appears to be more random than the other three data sets. 

Of course, to do all of this you will need to take subsets of your data and scan each section x points at a time.  The number of points to look at is going to vary based on the rate that you are acquiring data, how fast you expect your subjects to walk, etc.  I would use something similar to this example in order to count the number of peaks from walking which would tell you the number of steps.

I wish you the best of luck on your project!

Regards, 

Trey C.

[MJG3]

Trey,

Thanks for the direction. It is helpful to know that DIAdem is capable of the analysis.

Thanks,

MJG3