Tuesday, May 26, 2009

early tests on feedback and veering - charts



I have been asking students at my agency who require improvement in the veering to allow me to record the numbers on their normal practice, and the trend in their veering behavior after feedback. One student has had the most (n=34) treatments of feedback (indications to move left or right) and engaged in 34 trials where the number of paces he could walk before contacting a wall (in an 8 foot wide hallway) with his cane. Above is a graph, the x axis being the number of trials and the y axis being the pace walked. There is a trend toward improvement but not dramatic.

The graph above shows the results of trials and the average paces walked (6 trips on the path) without touching a wall. The initial trial is an average about 27 paces, the least number of paces is at about the 24th trials with about 18 paces walked, and the two longest average distances walked without touching a wall are at about the 28th and 34th trials. The trend is for less veering.

I reviewed Guth’s article and his email comments to me. He saw improvement after 6 to 8 trials of 15 treatments each. This means after 120 to 160 individual treatments saw recognizable improvement, so our results seem okay. I infer one major issue from this. We need to figure out how to do many more trials than the originally planned 8 to 20. We must do 60 or more at a minimum with each subject to see improvements.

I’ll continue to record the data on my students, doing about 8 treatments and two trials a week. The preliminary numbers look okay and that is the good news.

Here is a chart on the second student with simlar results - initial paces about 22, and the most paces walked after the 7th and 13th (last) trials:

Arc Width Measurement Demonstration


As part of the last Friday's project meeting, we demonstrated the latest arc width measurement software application. This application analyzes raw acceleration data from a Wii Remote to compute timing, velocity, and distance for each arc. The resulting distance measurements, together with user set thresholds, are used to generate spoken audio feedback.

On-screen slider controls are used to set thresholds for determining whether an arc is too wide, too narrow, or in the correct range. Other controls are used to set feedback options allowing the user to specify whether feedback should or should not be given when an arc is correct or is too wide or narrow.

Ellen Rubin, Gene Bourquin, and Steven Landau tested the system at the Touch Graphics offices and results were generally positive though the methods used to compute arc width need additional work. Currently, the biggest shortcoming is in the math used to double-integrate instantaneous acceleration to produce distance. Any "roll" in the cane which moves the Wii Remote's Y-Z plane off of alignment with gravity affect the calculated distance traveled along the Remote's X axis.

Future work will concentrate on better using static acceleration readings and an external reference to the light strip to improve arc width distance calculations.