Time-varying coherent AM–FM demodulation and denoising of acoustic signals

Abstract

Noise removal via linear time-invariant (LTI) filtering is most effective when the signal and noise spectra have minimal overlap in frequency. In particular, it can be difficult to extract, via LTI filtering, broadband signals from broadband noise, because often their spectra overlap. However, many broadband signals are locally narrow band (e.g., AM–FM signals with large FM and moderate to small AM), and this characteristic can be exploited to improve noise suppression for such signals. We present a method for extracting locally narrow band signals from broadband noise, based on an AM–FM decomposition of the signal and time-varying filtering. The center frequency and passband of a linear time-varying filter are determined from estimates of the instantaneous frequency and instantaneous bandwidth of the signal. Results on both synthetic signals and recorded whale sounds in ambient noise demonstrate a significant improvement in SNR compared to LTI-based filtering. [Supported by ONR Grant N00014-98-1-0680.]