next up previous
Next: Reducing Information Overload Up: StartleCam: A Cybernetic Wearable Previous: Experimental Validation

The Robustness of the Algorithm for Alternate Sensor Placements

To make such a system truly wearable it should be unobtrusive and fit comfortably into normally worn clothing, jewelry, and accessories. In earlier work, we explored the option of placing the electrodes on the toes and measuring skin conductivity across the sole of the foot [PH97]. We evaluated the startle detection algorithm on these measurements, to see if the startle response could be detected as easily from the foot as from the hand. An example of the startle detection algorithm working on simultaneous data from the foot and the hand is shown in Figure 7. The algorithm detects startles in the foot's signal that match the startles detected from the hand's signal. Even more placement options should yield similar results, since the eccrine sweat glands that produce this response cover the entire body. Because the response uses a bulk measurement conductivity, measuring a large area of skin can compensate to the smaller concentration of glands in areas other than the palms of the hands and the soles of the feet. Other researchers in our group are currently exploring skin conductance readings taken from the large area of the back in electrodes embedded into a shirt worn by a conductor of a symphony[MP98].

The skin conductance response is one of the most reliable physiological signals[SF90] and should be the easiest measurement to incorporate into clothing. The startle detection algorithm detects only sharp changes in the skin's conductivity response and is not triggered by slow changes in baseline. Therefore, sweat gland activation for thermal regulation or baseline drift from deterioration in the conductance of the electrodes over time will not generate false startle responses.

  
Figure: The StartleCam algorithm can detect responses in skin conductivity signals from placement on either the hand or the foot. Figure (a) shows the detected startles on a signal from the hand and Figure (b) shows the algorithm working on a signal taken from the foot at the same time.
\begin{figure}\centerline{\psfig{figure=/v/projects/AC/jen-pics/hfscamdata.eps,width=85truemm}}
\end{figure}


  
Figure: StartleCam creates a composite image from a buffer of digital images recently captured by the wearer's camera. In this sequence, the wearer was surprised by a question from a friend. The sequence shows images captured as the wearer turned to address her friend.
\begin{figure*}\centerline{\psfig{figure=/v/projects/AC/jen-pics/scam52.ps,width=145truemm}}
\end{figure*}


next up previous
Next: Reducing Information Overload Up: StartleCam: A Cybernetic Wearable Previous: Experimental Validation
Jennifer Healey - fenn@media.mit.edu
1999-02-12