Matrix analysis and applied linear algebra
Matrix analysis and applied linear algebra
Multichannel Blind Source Separation Using Convolution Kernel Compensation
IEEE Transactions on Signal Processing
| Hi-index | 0.00 |
This paper studies a new version of a sequential decomposition method suitable for real-time separation of linear mixtures of finite-length signals. The signals are modelled as outputs of a multiple-input multiple-output (MIMO) model with finite channel responses and trains of delta pulses as inputs. Our decomposition method compensates the channel responses and aims to reconstruct the input pulse trains directly as the blocks of samples of observed signals become available. This new block-sequential version of the decomposition is compared with already developed sequential version of the same decomposition algorithm in terms of speed and performance. Tests on simulated signals show the novel version is less computationally intensive than the previous one and exhibits comparable accuracy of reconstructed trains of delta pulses.