BLIND SOURCE SEPARATION WITH CONVOLUTIVE NOISE CANCELLATION

On-line adaptive learning algorithms for cancellation of additive, con volutive noise from linear mixtures of sources with a simultaneous bli nd source separation are developed. Associated neural network architec tures are proposed. A simple convolutive noise model is assumed, i.e. the unknown additive noise in each channel is a (FIR) filtering versio n of environmental noise, where some convolutive reference noise is me asurable. Two approaches are considered: in the first, the noise is ca ncelled from the linear mixture of source signals as pre-processing, a fter that the source signals are separated; in the second, both source separation and additive noise cancellation are performed simultaneous ly. Both steps consist of adaptive learning processes. By computer sim ulation experiments, it was found that the first approach is applicabl e for a large amount of noise, whereas in the second approach, a consi derable increase of the convergence speed of the separation process ca n be achieved Performance and validity of the proposed approaches are demonstrated by extensive computer simulations.