HOW DOES SETIEasy WORK?

A publication describing the basic spectral analysis algorithm used to search through the signals in the frequency domain is available as a 1.1 Mb PDF file.

SETIEASY is a computer program to detect unusual small signals in a large background of radio noise. Through the analysis of microwave radio signals collected from space beyond our solar system, SETIEASY sifts through noise for signals that might be evidence of intelligent extraterrestrial life. The program is text-only with minimal output to the screen, compiles easily using a standard g++ compiler or Borland C++v5.02 compiler, and uses only 7 megabytes of RAM. SETIEASY's input is a stream of floating-point values from a Project Argus radiotelescope that are saved as WAV sound files. SETIEASY takes a series of data windows from the input stream, beginning from a fixed position in the data file, and extending the windows exponentially to include progressively more data. The smallest data window size is about 8 seconds and the largest is over 120 seconds. The files are typically 15 minutes or one hour long. The computer program must first extract the time-domain floating point values from the data file and compute the fast Fourier transform of the extracted set . The DFFTs of the data are the input for the Replica function, which estimates the multidimensional variance of the calibration set by creating 100 bootstrap samples of the post-DFFT data set. Bootstrap samples are data sets drawn randomly and with replacement from the input DFFT data set. The creation of repeated bootstrap samples is Monte Carlo integration of the bootstrap distribution function. Each time a bootstrap sample is created, the hyper-dimensional coordinates of the center of the DFFT data set are computed. In the case of SETIEASY, the DFFT graph is expressed in an 8192-dimensional hyperspace, with the displacement of a spectral point on a frequency axis denoting the intensity of the signal observed at that frequency.. For example, three values (x, y, z) are needed to graph a spectrum with three frequencies as a point in 3-dimensional space, so 8192 values (x, y, z, a, b, c, d, e, f, g, h, i, j, k, ...) are needed to graph a spectrum with 8192 frequencies as a point in 8192-dimensional hyperspace. The Replica function is applied twice to two pieces of the DFFT data set, a calibration set and a test set. At this point, the training set and the test set are each 100 x 8192 element arrays of values. The centers of each data set are computed to produce two 8192-element arrays. Using the spectra and their centers, Minkowski distances are computed between the center of the calibration set and the points in the calibration set, and between the center of the calibration set and the points in the test set. Once this reduction of spectra in hyperspace to simple distances is done, the correlation between the calibration set distance integration and the test set distance integration is calculated. The calibration set correlation is calculated for bootstrap replications of itself for the first block of time in the data stream, and then more times for between subsequent blocks of time and the first block of time. The block of time is variable in length, and currently ranges from 2 seconds to 64 seconds, testing all powers of two in between these limits automatically. Currently, the confidence limits for SETIEASY are set to accept over 99% of the values as statistically random, and the remaining less than 1% are flagged as unusual and worthy of further investigation. Although this setting calls attention to more noise than true extra-terrestrial signals of intelligent origin, there is a greater chance of finding weak signals from intelligent life if any slightly unusual signals are flagged for closer examination.
Note: Getting signals flagged as Unusual with a message to mail the output file to us does not indicate you have a "hit!" Any interference or burst of noise that appears in the first model block can cause all subsequent blocks to "appear" unusual. The shorter the integration time, the more likely many noises are to trigger the flag to look at the data more closely. We really need BETA testers to try out the program on their PC's, workstations, etc. SETIEASY is a BETA version. We hope you will help us complete the debugging that would make it possible to distribute to any system architecture or operating system. We have made the source code and the compiled version of Setieasy for Win95/98/NT available. A readme text is also available. All beta testers are welcome to report bugs on the discussion groups or by email.

DISCLAIMER: The authors of this product are students and can accept no responsibility for damages resulting from the use of this product and make no warranty or representation, either express or implied, including but not limited to, any implied warranty of merchantability or fitness for a particular purpose. This product is provided as is, and you, its user, assume all risks when using it. Robb Samuell - March 21, 2000