Multiple-channel active noise control systems

Includes bibliographical references (pages [80]-81)In this thesis, we focus on a real-time implementation of multiple-channel active noise control algorithms for noise canceling in a large area. The algorithms considered include multiple-channel filtered-X least mean square (MFXLMS), multiple-channel feedback prediction filtered-X least mean square (MFBLMS), and multiple-channel hybrid filtered-X least mean square (MHFLMS). The first two algorithms are discussed in detail and the last algorithm will be introduced in this thesis. The MHFLMS algorithm, which basically is an integration of MFXLMS and MFBLMS algorithms, can cancel both narrowband and broadband noises. The broadband noises to be canceled must be correlated with the reference signal picked up by the reference sensors. But the narrowband noises to be canceled may not be correlated with the reference broadband noise signal. As shown by the results of computer simulations, MHFLMS algorithm can be very effective in certain complicated noise environment. In this research, firstly computer simulation was used to compare the performance of all three multiple-channel active noise control algorithms. Many noise models were considered in the simulation to fully explore variations of the algorithm performance under different conditions. Secondly, a real-time implementation, carried out under a laboratory environment, was used to evaluate the performance of the multiple-channel active noise control algorithms. The real-time implementation used a TMS320C31 Quad Processor Board from Spectrum Signal Processing Inc. Finally, extensive real-time field tests were conducted in a large manufacturing plant to demonstrate the applicability of the above algorithms to practical applications. Simulation and experiment results are presented and discussed in this thesis and future research directions are suggested.M.S. (Master of Science