Presentation at the Computational Neuroscience Meeting (CNS)

The 25th annual Computational Neuroscience Meeting (CNS 2016) will be held July 2-7th, Jeju Island, South Korea. Minho Song will present a poster about Sufficient sampling frequencies for fast hand motion tracking.

Summary

When tracking fine motor behaviors in human body parts, passive marker-based tracking is one of the best-suited methods not only because of its high spatial precision and temporal resolution, but also allowing high degrees-of-freedom [1].  However, the passive marker approach suffers from identity confusion problem (Fig 1A) between the markers. As the speed of motion increases, sufficient sampling rate is required to avoid the problem. In a recent study [2], we reported that the problem still occurs even with the sampling rate significantly higher than the Nyquist sampling rate. The study suggested a sampling rate criterion to avoid identity problem for the worst-case condition.

In this poster, the confusion problem is tested in more realistic human motor control behavior. Grids of 3x3 markers with different distances (1, 1.5 and 2 cm) were attached to a skilled piano player’s right hand (Fig. 1B). The experimental task was repeated right-hand alternative keystrokes between D#5 and D#7 (two octave) with a tempo of 176 bpm for 10 seconds. This is an excerpt from Liszt’s La Campanella, which requires fast horizontal jump of the right-hand. These motions were recorded with 7 optical motion capture cameras (Qualisys Ltd. Oqus 400) changing the sampling rates from 50Hz to 200Hz. The maximum frequency components of these hand movements were lower than 8Hz.  

The probability of successful tracking is measured by counting the number of successful repetition of the center marker (Fig 1C). Estimated required sampling rates for successful tracking (where the probabilities reach 100%) were 101Hz, 137Hz, and 181 Hz (fitted to piecewise linear functions by expectation maximization). The theoretically predicted values are 176Hz, 235Hz, and 353 Hz [2].

We found that the required sampling rates are lower than the theoretical criterion. This is because the theoretical prediction was developed to avoid the worst case where marker trajectories overlap from perfect periodic motion; not realistic for human movement, which has variability. Our results show that in practical situations involving human movements, the sampling criterion can be weakened considerably. But, it should be note that a motion slower than 10Hz still requires more than 100Hz, which far exceeds the Nyquist sampling rate.  

Figure 1. Experimental design and the result. A. Marker confusion. Grey dots are markers. ,  are the distances between markers.  is the sampling latency.  is speed of marker. Green lines show the markers, which identified as same. left correct identification right example of marker confusion. B: experimental set up. Red dots are keys to press by the thumb and the little finger during repeats. C. The probabilities of continuous marker identification.

 

References

[1] Guerra-Filho G. Optical Motion Capture: Theory and Implementation. Journal of Theoretical and Applied Informatics. 2005; 12(2):61-89

[2] Song M-H, Godøy RI (2016) How Fast Is Your Body Motion? Determining a Sufficient Frame Rate for an Optical Motion Tracking System Using Passive Markers. PLoS ONE 11(3).

Published June 23, 2016 11:35 AM - Last modified Dec. 6, 2018 5:21 PM