Subspace Detection in Subspace Interference for Underwater Acoustics

A process called subspace detection in subspace interference SDSI was developed. The goal of this work was to produce an algorithm that could provide a fast estimate of the noise or interference subspace in beamformed data. The SDSI process involved separating a signal into two perpendicular subspaces, one containing noise or interference and one containing the signal alone. Detection would then be done in the subspace containing the signal only, improving detection and helping defeat countermeasures. To make the algorithm adaptive, the functions used to form the subspaces were taken from the time-frequency analysis waveforms used in the wavelet and local cosine transforms to produce good subspaces that would represent the noise well but not the signal that it is desired to detect. Projecting data into a space that is perpendicular to the noise subspace removes any of that noise in the data. The projection is performed on the matched filter for the desired signal to be detected. This should produce a modified matched filter that does not react to noise present in the data but still detects the signal. However, the output of the modified matched filter was generally not much better than that of the original matched filter. Noise couldnt be significantly separated without degrading the signal. Future work should focus on finding a better choice of bases that separate the noise and the signal better than the local cosine.