The SETI@Home Analysis Algorithm
From Unofficial BOINC Wiki
[edit] General
The SETI@Home program (Science Application) performs a variety of digital signal processing operations on radio frequency data. The algorithm consists of two parts:
- transform the input signal by smoothing, chirping and conversion into power spectra, and
- analyse these power spectra by:
- finding spikes in individual spectra,
- analyse the evolution of the power of a particular frequency over time by look-
ing for Gaussians, Pulses and Triplets. This is referred to as power over time analysis.
This description only contains details of how the algorithm works, not justification for why this would be useful in detecting an extraterrestrial signal.
[edit] Transforming The Signal
The input signal X consists of N samples represented by complex numbers:
- <math>\X = {x0 , x1 , . . . , xN −1 } xj \in \mathbb{C}</math>
- <math>\X = {x0 , x1 , . . . , xN −1 } xj \in \mathbb{C}</math>.
In particular, the SETI receiver performs -bit complex sampling, i.e. each sample of the signal is represented by 2 bits of data. This means each sample can take one of four possible values: (0, 0), (0, 1), (1, 0) or (1, 1).
This 2-bit format is not useful for actual processing. Before analysis, the data must be unpacked to the floating-point format used by computers. In practice, this is the IEEE single precision format, where each number is represented by 32 bits (4 bytes). Therefore, each sample is converted to two such numbers, for a total of 64 bits (8 bytes) per sample, using the following conversion:
(0, 0) → −1 − i (0, 1) → −1 + i
(1, 0) → 1 − i (1, 1) → 1 + i
