Forensic audio recordings may contain undesired noise that can impair source identification, speech recognition, andother audio processing requirements. In this paper several custom analysis/synthesis algorithms are presented based ona time-varying spectral representation of the noisy signal. The enhancement process adapts to the instantaneous signalbehavior and alters the noisy signal so that the enhanced output signal is higher in quality than the unprocessed noisyinput signal. Nonlinear and time-varying filters operate on the spectral representation in order to retain features that areattributable to the desired signal, such as human speech, while removing the features that are more likely to be due tothe noise contamination.
展开▼