5.3  Data Extraction and Handling

The first segments of data that will be analyzed are those sections of data that correspond with the appearance of a visual stimulus (across all physiologic channels).  The duration of the visual stimulus in this experiment was 2.00 seconds or 1900 samples.  Extracted data segments were chosen to be 2.156 seconds in duration (2048 samples) which covers the entire length of the visual stimulus (1900 samples) with an extra 0.156sec. (148 samples) at the end so that the (double-precision) radix-2 FFT algorithm can be used without zero-padding the data.

Figure 5.3.1 shows a visual representation of the above outlined procedure and shows which portions of the physiologic data will be extracted relative to the occurrence of the two example visual events marked for extraction.  Extracted data is indicated by the shaded regions corresponding to the occurrence of visual events.

Figure 5.3.1  Data Extraction Example

Extraction software

Since each of the tests performed result in several hundred megabytes of data spanning multiple data files, software was produced to handle all of the data extraction which is necessary in every phase of the data analysis.  Typically, the desired starting and ending sample are specified (usually obtained from the automated test reports), and the software extracts the appropriate data across all channels into a separately marked data file.  In addition, an information file is generated for each extracted segment which contains specific information related to that particular segment, and more detailed information is written to a running logfile which can be used to verify the origin and parameters used in extracting any data segment.  The information files generated for each piece of data which is extracted are produced in machine readable form and is used by all data processing software throughout the experimental analysis.

When collecting EEG data, on average, it is usually necessary to change the gain settings on the amplifiers at various times throughout the tests to either increase the signal amplitude to fill the full voltage span (thereby maximizing the A/D resolution) or decreasing the amplitude to prevent saturation at the amplifiers or data acquisition equipment for each of the channels.  The software used to extract the data segments adjusts the data to the (calibrated) amplifications used when collecting the raw EEG signal and scales the output amplitude to micro-volts (other scaling options are also available).  This allows the direct comparison of EEG segments which were collected at different amplifications during a test.

Figure 5.3.2 shows an example of data from this experiment which has been extracted for subsequent analysis.  The first 4 rows of the figure correspond to the EEG channels (Frontal, Central, Parietal, and Occipital), row 5 is the vertical EOG channel showing a typical eyeblink (negative deflection is eye closure and positive deflection is eye opening), and row 6 is the ambient electrical noise channel.  All channels are displayed on a micro-volt scale.

Figure 5.3.2  Data Segment Example (2048 pts./channel)

<PREVIOUS

NEXT>