Research Of Signal Reconstruction And Information Detecting Based On ICA

Independent Component Analysis (ICA) is a blind source separation algorithm, the model can be isolated through its independent data components. Signal reconstruction, ICA save the law of the development process of reconstruction, only the algorithm model of the hidden components of the signal tap, and the laws of ICA reconstruction model. With the traditional methods of signal reconstruction without a priori knowledge, to broaden the scope of the reconstruction, simplifies the process of reconstruction. In Information Retrieval, ICA without the development of classification standards, the direct use of different types of statistical independence of signals in the separation model. Therefore, ICA applications in these areas of theoretical study of novel, practical features. In this paper, using ICA techniques, signal reconstruction and the application of information retrieval problems in the study, the main work and conclusions are as follows:(1) An overview of domestic and foreign ICA main results of research and development trends, focusing on a number of typical characteristics of the ICA algorithm and through experiments on the separation performance of these algorithms are compared.(2) The use of the ICA and the kernel function of combining independent component analysis algorithm for nuclear (KICA), and applied to Face Recognition. The results show that the new algorithm has very good recognition results.(3) In order to meet the reconstruction of the weight of the order and magnitude of the request will be improved KICA, the formation of an improved algorithm for independent component analysis of nuclear (IKICA), through the successful implementation of the algorithm is a three-dimensional face image reconstruction.(4) An analysis of hyperspectral images of the model, the fast independent component analysis algorithm (FastICA) adjusted to the different features of Hyperspectral Image Segmentation region. The results show that the unsupervised learning case, by virtue of ICA algorithm can be completed in different regions of the partition.