Development of Improved Active Noise Control Systems and their Real-Time Evaluation

Abstract

Noise pollution and its control have been an important topic of research due to rapid industrial and residential growth and stricter standards. Passive noise control techniques using enclosures, barriers and silencers (mufflers) are generally used to control the noise at the source or at the medium. However, with all such technology available today, low-frequency noises below 500 Hz are not effectively controllable by passive means. Therefore, an active technique which uses the principle of superposition has been gaining interest among researchers as an important topic of research. Active noise control (ANC) is a technique by which anti-noise (equal magnitude and opposite phase) is generated to attenuate the offending noise. Even if there are many algorithms and schemes used to make ANC usable, still due to certain issues, the applications of ANC are very limited. Some of these issues linked with active headrest, nonlinear noise process, virtual microphone control and mixture of chaotic and tonal noise are dealt in this thesis.
The active headrest is one of the important applications of ANC. These headrests can be used to design acoustic comfort zones in cars, airplanes and industrial areas etc. There are many implementation schemes for active headrests which provide a good amount of noise reduction as long as the head is at rest. However, when the head is moved even slightly, the noise is felt to increase. Therefore, a need is felt to further analyze the existing active headrest system and then propose improved algorithm and implementation scheme such that the performance of active headrest is immune to head movement. In addition to this, it was found that there is a scope for improving the performance of ANC system for nonlinear noise processes and for noise which contains both tonal and chaotic components.
The performance immunity of active headrest to head movement is newly proposed in this thesis using head mounted movable microphones in an active headrest system. Such concept is utilized to design a hat with two microphones kept near to the ears. The ability of such a scheme is tested to control both linear and nonlinear noise processes. For linear noise process, filtered-X least mean square algorithm with a common set of secondary paths and moving error microphones are used. Nonlinear noise processes are controlled by a nonlinear filter bank based ANC algorithm for headrest application. To further improve the noise control performance in an active headrest, two modified virtual control algorithms are proposed: one for linear noise process and the other for the nonlinear noise process. Through real-time experiments, it was found that the proposed virtual control strategies are best suited for active headrest application. Another improved ANC algorithm is developed to have better performance when the noise contains both tonal and chaotic components.
It was found that with the help of head mounted moving microphones, the noise in a headrest can be actively controlled for any head positions in a headrest. The nonlinear ANC algorithms can also be used with moving error microphones for headrest application. The conventional virtual microphone control algorithm is not suitable with moving microphone, whereas the proposed modified virtual ANC algorithm works seamlessly for all head positions in a headrest system. The two proposed modified virtual microphone control based ANC algorithms one for linear noise and other for nonlinear noise processes are superior to their physical microphone based control strategies. In addition to the work related to the active headrest, the algorithm that has been proposed to tackle the noise which contains both tonal and chaotic components is an efficient one. The results show that if the tonal and chaotic noise components are first separated, and anti-noise is generated by a separate controller, the noise control performance is better than a common controller using the whole noise. The proposed algorithm adaptively separates both the noise components and generates appropriate anti-noise for control purpose.