Case study of principal component inverse and cross spectral metric for low rank interference adaptation

Freburger, B.E.; Tufts, D.W.

doi:10.1109/icassp.1998.681528

Cited by 5 publications

(6 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although the active sonar data we use here is after beamforming and it doesn't meet the classical BSS model, real sea experiment proves that the approximation to the BSS model though backward matrix is effective, and our method is valid in nulling of reverberation with active sonar data. There are three main differences between our method and classical manner in nulling of reverberation in [4,5] .One difference is length of active sonar data used in the algorithm. The classical detection is nearly within the duration of transmitted signal, and our algorithm works on the data of several, even dozens of durations, and the second is that we make use of the colored property, not whiten the data, and the third is that our method does not need more important information except active sonar data, but the method in [4,5] can't process the active sonar data with the prior information in this paper.…”

Section: Resultsmentioning

confidence: 99%

“…There are three main differences between our method and classical manner in nulling of reverberation in [4,5] .One difference is length of active sonar data used in the algorithm. The classical detection is nearly within the duration of transmitted signal, and our algorithm works on the data of several, even dozens of durations, and the second is that we make use of the colored property, not whiten the data, and the third is that our method does not need more important information except active sonar data, but the method in [4,5] can't process the active sonar data with the prior information in this paper. This contribution has demonstrated the effectiveness of using BSS on active sonar data for nulling of reverberation.…”

Section: Resultsmentioning

confidence: 99%

“…As Neumann and Krawczyk [8] have pointed out the forward matrix model used is not a principal issue of the algorithm of PCI, and any other forward model can be used. So we generate the backward data matrix [4,5,8] …”

Section: Backward Matrix Of Active Sonar Data Formentioning

confidence: 99%

“…Here, the length of ) (t x k is more or less dozens of samples shorter than ) (t d . In [4,5], the size of forward-backward matrix is small, and usually it is one eighth to one sixth of one duration. But the size of our backward matrix is much larger because the data that we use to generate the matrix is much longer.…”

Section: Experiments and Analysismentioning

confidence: 99%

“…For the latter factor, some important prior knowledge, for example the target power should be provided [4], and usually it is assumed that reverberation is a sum of returns issued from the transmitted signal. Principal Component Inverse (PCI) [4,5] has been applied to separate reverberation and target echo with the forward matrix that is generated from active sonar data. These considerations are needed to detect the target echo accurately in the reverberation dominant environment, especially in shallow water.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Blind signal separation and reverberation cancelling with active sonar data

Cong¹,

Hu²,

Shi³

et al.

Proceedings of the Eighth International Symposium on Signal Processing and Its Applications, 2005.

View full text Add to dashboard Cite

With easily available prior knowledge, we explore a new approach for nulling of reverberation through separating active sonar data into reverberation and moving target echo. From the perspective of signal processing, the method, as outlined below, is based on active sonar data properties and the process of how reverberation is formulated. It is assumed that active sonar data can be separated into two parts that are target echo signal and reverberation. First, we generate a backward matrix from active sonar data to approximate the classical blind signal separation model. Second, after blind signal separation is performed on the backward matrix, separation is done. Third, the estimated signal with the largest kurtosis is selected as the desired target echo signal and the reverberation is cancelled at the same time. Fourth, the human hearing feature helps the process to be adaptive. The real sea experiment proves the method to be effective. 0-7803-9243

show abstract