Sub-band, dual-channel adaptive noise cancellation using normalised LMS

An adaptive noise cancellation scheme for speech processing is proposed. In this, the adaptive filters are implemented in frequency-limited sub-bands, based on a simplified model of the human cochlea. A modification to the basic LMS structure is introduced which guarantees stability and convergence at all times. This modification, a non-recursive normalisation, is used both in a wideband and in a sub-band implementation of the scheme. The effect of this normalisation on the quality of the processed speech is to cause the speech to be distorted, showing that there is no benefit to using normalised LMS in a sub-band scheme, whether the application uses classical or intermittent noise cancellation.