Preliminary data from the first pilot experiments are presented to illustrate features of interest. Figure 11 shows the data collected by four sensors during two hours and twenty minutes of driving. Some events which were recorded from the audio record were the driver merging onto the highway at 31 minutes, the driver establishing cruise control from 38 to 47 minutes and the driver turning off the highway at 122 minutes. Looking across multiple sensor data suggests that events which involve both physical and mental stress such as turning can be distinguished from events which involve primarily mental stress such as merging or little stress such as driving under cruise control.
One confounding difficulty, however, is that the frequency of responses also appears to vary for the same driver on different days. Figure 12 shows the same driver's response to similar driving tasks. To account for variations due to gel and sensor placement the data is normalized by subtracting off the minimum and dividing by the range (maximum - minimum)[SF90]. In the first experiment, many responses occur, but in second experiment shown, only one strong reaction occurs when the car nearly collides. However, the strength of this reaction, may have also skewed the normalization process, by providing a larger range than the first experiment.
 |