Signal Processing Aspects of Signal Detection Masking and Noise Suppression

Acoustic signal extraction and identi cation in the underwater environment is best achieved by adaptive methods as the signals encountered are generally non-stationary and corrupted by unpredictable noise sources, such as man-made noise, biological and seismic noises. While classical methods often fail in such an environment, the recent use of multiresolution methods like the adaptive wavelet transform and its dual, the cosine packet transform, provides a promising alternative. This paper introduces three applications where wavelet and cosine transforms are used for denoising and signal identi cation. Signal decomposition on these two sets of functions provide di erent representations, which are each speci c to a certain noise type. The third application achieves background pink noise ltering along with a high signal compression rate, which can be used to optimize signal identi cation.