Computing multiple spectrum records from a signal

Here is an example of using the component software to compute the time-varying spectrum of a single EKG signal. In this example, the average signal for electrode location V(3, 3) is used.

To compute a time-varying spectrum, a moving subset of the signal is extracted. In this example, each subset is 1/8 of a second, or 32 points, long. A DFT is performed on each of the subsets to produce a spectrum record showing the frequency components during that particular 1/8th second window of the signal. For example, the subset of the signal containing the RS "wave" of the EKG has the following corresponding spectrum magnitude for frequencies up to 50 Hz. The peak of this particular spectrum record is at about 15 Hz.

All of the overlapping spectrum records can be assembled into a 2d dataset by a FIFO component which has been configured to output all of its slots at the same time. This 2d dataset can be plotted as a surface or image, showing the change in the signal's spectrum over its entire duration (in this case, 1 second). In this plot, time runs along the -Y axis.

Although this analysis has been performed for only 1 out of the 25 signals in the EKG grid, it could be performed for all of them and displayed as a set of 25 surfaces which vary with respect to electrode location. This would be an alternative to displaying the spectrum as a frequency "slice" for all 25 coordinates, as was done in the previous page.


©Copyright Sky Coyote, 2001-2002.