Multichannel Adaptive Stochastic Filtering for Active Noise Control in Personal Computers

In this paper, the feasibility of using active noise control inside personal computers is demonstrated by applying an adaptive filter to Hewlett-Packard's Blackbird 002 Gaming PC. Conditions relating the control signals and the sound heard by an external listener are presented as a means of evaluating the hardware design before applying the adaptive filter. A multichannel stochastic adaptive filter is derived by vectorizing the equations and applying gradient descent and Newton's method. For implementation purposes, data-based approximations to the gradient and hessian are provided. It is shown that a 6dB reduction in sound pressure level is obtainable by adding reference microphones, error microphones, a speaker, and some acoustical foam into the system even though the length of the noise cancelation system is small, approx. 6, compared to the fundamental wavelength of the noise, approx 23.