Abstract:In this paper, a new method for extracting interested source signals in the under-determined anechoic mixtures is presented. The independent component analysis with reference (ICA-R) algorithm is used to extract the non-stationary signals from the prior azimuth information of the interested source signals (including the arrival angles and the distances between the source signals and the array elements). Firstly, construct the initial weight vector by using the prior azimuth information of the interested source signal. The initial weight vector can be utilized to obtain a reference signal for the interested source signal. In order to make the weight vector converges to the optimal solutions quickly and accurately, the closeness measurement function of the interested source signal is introduced on the basis of the complex FastICA algorithm. Secondly, we take the weight of the closeness measurement function and the complex FastICA optimization function as a new optimization function. Finally, the weight vector is updated by the iterations of the proposed optimization function. When the algorithm converges or reaches the maximum number of iterations, we can obtain the final weight vector and extract the interested source signal. The proposed algorithm can extract the interested source signals efficiently and quickly by utilizing the prior azimuth information of the interested source signals. The results of the simulation experiments show that the proposed algorithm can extract the interested source signals very well, even if there is a great error between prior azimuth information and real azimuth information.

Combining independent component analysis with geometric information and its application to speech processing

Overcomplete ICA with a Geometric Algorithm

Overcomplete BSS with a Geometric Algorithm

Generalizing Geometric ICA to Nonlinear Settings

A Multimodal Approach for Frequency Domain Independent Component Analysis with Geometrically-Based Initialization.

Geometric Overcomplete ICA

New Geometric ICA Approach for Blind Source Separation

Blind Source Separation of Underwater Acoustic Signal by Use of Negentropy-Based Fast ICA Algorithm

An MRF-ICA based algorithm for image separation

An improved geometric overcomplete blind source separation algorithm

Independent Component Analysis based speech separation using geometric k<sup>th</sup> nearest neighbor entropy estimator

Approaches and applications of semi-blind signal extraction for communication signals based on constrained independent component analysis: The complex case

A Blind Source Extraction Algorithm Based on ICA-R for Underdetermined Anechoic Mixtures

Constrained Complex-Valued Ica Without Permutation Ambiguity Based On Negentropy Maximization

Blind Extraction of Independent Signals from Their Linear Mixtures

Oblique projectors-based blind source separation using information maximization principle

Independent Component Analysis and Its Application to Overlapping Speech Pitch Extraction and Speech Enhancement

Approach Based on ICA and GCA to Identify Field Mixed Acoustic Objects

Speech blind signal separation with FastICA and Markov Chain combination

Mutual Information Based Approach for Nonnegative Independent Component Analysis

Underdetermined Blind Radar Signal Separation Based on ICA