Real-time active noise control and speech enhancement for functional MRI

High level of broadband acoustic noise, created by the gradient coils in a 3-Tesla functional Magnetic Resonance Imaging (fMRI) scanner contaminates the speech communication between the patient and the researcher. Speech recovery with regard to fMRI acoustic noise is of great interest to experimenters who wish to obtain verbal responses from the subjects. Recovery of speech from the noisy recordings is challenging due to the broadband and dynamic nature of the fMRI noise. The noise level is much higher than the speech level making speech enhancement a challenging problem. This thesis presents the development and the performance analysis of multi-band and sub-band based adaptive speech enhancement methods suitable for recovering speech corrupted by fMRI noise.; A LabVIEW based Real-time Active Noise Control (ANC) system that is designed to cancel acoustic noise at high sound levels is also presented. The ANC system is evaluated on a realistic test-bed mimicking an actual fMRI bore.