Applied Methods to Combat Noise in Human Communication

Abstract: Acoustic noise has a pronounced negative impact on human vocal communication and user comfort. Interfering noise that is transmitted between communicating users may not only disturb telephone conversation and supporting technical mechanisms (such as speech coders), but may also strain and eventually damage hearing and vocal organs of the users. Furthermore, a user subjected to consistently high levels of noise (such as those emitted by heavy machinery or electrical hand-held tools) may suffer discomfort and health risks. In many cases it is desirable to reduce the level of noise transmitted over a network, as well as controlling and minimizing the level of noise that the local user is subjected to through the utilization of methods for active noise control. This thesis provides applied methods to combat noise in human communication. The active hearing defender is one key application for which the introduced methods may be suitable. The first three parts of this thesis comprise methods for noise reduction in user-to-user vocal communication. In particular, blind methods are investigated. As opposed to conventional (non-blind) methods, blind methods are stand alone and do not require a priori information or knowledge of the spatiotemporal environment. A method based on statistical kurtosis is provided, in which an adaptive, blind subband beamformer is derived and evaluated. The performance of this approach supports speech enhancement in a wide range of applications. A new, low complexity method for blind beamforming is introduced, in which several single channel blind speech enhancers are linearly combined. The nonlinear nature of the blind speech enhancers mean that speech sources add coherently, i.e. performing blind beamforming. Important aspects of implementing a single channel blind speech enhancer in analog, digital and hybrid (mixed analog-digital) hardware are also analyzed. The final section of this thesis outlines an application for active noise control for the purpose of hearing protection. A low-power fixed point digital signal processor is used as an implementation platform as it supports battery powered apparatus. The broad band noise reduction is 20 dB to 30 dB and the tonal interference rejection is 60 dB.

  CLICK HERE TO DOWNLOAD THE WHOLE DISSERTATION. (in PDF format)