| dc.description.abstract | During the work with this master's thesis, the possibilities of controlling the emitted sound from a person speaking by using active control, have been examined. The work included measurements of transfer functions on a setup with a head and torso simulator, acting as the noise source, an anti-noise loudspeaker, and microphones for detection and evaluation of performance. Based on the measured transfer functions, simulations were done to find the potential performance of the active control system by looking at the effects of modifying different parameters used in the design process. The vulnerability to deviations from the ideal case presented in the literature, and the impact of delays due to hardware necessary for a final implementation, was also investigated.
The results from the simulations were evaluated based on an estimate of the change in radiated power, and the reduction of the Articulation Index. The change of the radiated power was evaluated as a function of frequency. It proved to be possible to achieve a reduction in the entire frequency range from 200 to 2000 Hz, with a maximal damping exceeding 20 dB. Unfortunately, the control system did also increase the radiated power outside this frequency range with up to 5 dB. The active control did still give a net reduction of the average power, as well as a reduction of the Articulation Index of approximately 0.1, where 1.0 is the maximal reduction.
These results are distinct improvements from previous work, and it is assumed that there still are untested solutions that could improve the performance even further. The simulations do also indicate that modern hardware is good enough for such systems to be realized. | en |
