Abstract
Intelligibility of speech and other audio in various acoustic environments can be very low based on the level of noise present in that space. Often in order to compensate for the noise in acoustic environments a passive approach is taken, in which sound absorption and isolation techniques are implemented throughout the physical space. Although this passive approach can provide suitable levels of noise reduction for a reasonable range of frequencies, several active noise control approaches are proven to be more effective, especially at frequencies below 500 Hz. Active Noise Control utilizes adaptive filtering algorithms to achieve significant noise attenuation of lower frequency noise that significantly degrades the desired audio signal. The level of effectiveness of various adaptive filtering algorithms can determine the overall noise attenuation and the rate at which the system attenuates the noise in the given space. Noise source location, level, and resonant space may dictate the amount of degradation caused on the input signal. Some acoustic spaces in which Active Noise Cancellation (ANC) headsets are used causes increasing resonance and inconsistent changes in noise power and amplitude. Special consideration of these signal characteristics are used to determine the stationary level of the noise signal, which is an ideal measure for determination of an adaptive filtering algorithm. Most Active Noise Control systems currently utilize only versions of the Least Mean Square adaptive filtering algorithm. This research will focus on a selective algorithm that will utilize descriptors to identify differences in the noise signal, and determine an optimal algorithm to be used for the filtering of distinct portions of the signal that are identified. The Recursive Least Squares adaptive algorithm will be used as the primary means of filtering due to its fast rate of convergence and tracking performance. A selective switching mechanism will be used to insure the Least Mean Squares algorithm is initiated only when signal conditions dictate effective noise control can be performed using this algorithm. Several different noise source environments are tested in order to attain the best active noise control solution to be used.