MULTIPLE REFERENCE FEEDFORWARD ACTIVE NOISE CONTROL PART II: REFERENCE PREPROCESSING AND EXPERIMENTAL RESULTS

Abstract

Multiple reference active noise control (MRANC) has been applied to acoustical fields with multiple noise sources to achieve low frequency noise reduction. The traditional control configuration feeds each reference signal into a different control filter. This configuration has been widely adopted due to its potential performance in a general multiple noise source environment. However, it entails the problem of ill conditioning when the reference signals are correlated. In this paper time domain analysis has been carried out to investigate the problem of ill conditioning for MRANC. To cope with the problem of ill conditioning, a reference signal preprocessing step is added to the conventional active noise control process. This preprocessing step essentially constructs a new set of reference signals, which preserve all the information of the original reference, but are uncorrelated with each other. An adaptive decorrelation filter based on the Wiener filter theory and Gram–Schmidt orthogonalization theorem is constructed to implement the reference signal preprocessing stip. Experiments based on sound transmission through a vibrating plate have been conducted and the results presented are consistent with the theoretical analysis.