Application Of Dual-tree Complex Wavelet Transform In Weak Biomedical Signal Processing

Biomedical signals are generally the weak low-frequency and non-stationarysignals under strong noise background. They have a very important value for researchand clinical diagnostic of the human system state. Their processing methods are basedon the digital signal processing theory. In biomedical signal processing, the ability ofremoving the interference noise and extracting the feature information determineswhether people actually use these signals. So far, based on the Fourier transform, avariety of time-frequency signal processing methods have used to extract useful featureinformation for research and clinical diagnostic.The wavelet transform is a further development of the Fourier transform. Itinherited the localization thinking of the short-time Fourier transform. And also itovercomes the shortcomings of the time window size which does not vary withfrequency. It becomes a new milestone since the Fourier transform. It is known as a“mathematical microscope” in signal analysis. Due to the time-frequency analysis andmulti-resolution analysis, the discrete wavelet transform is widely used. Yet it hasseveral aspects disadvantages, including translation sensitivity, frequency aliasingphenomena and absence of directional selectivity. In signal processing, these defectswill lead to loss much detailed information or get inaccurate results.To solve the above problems, a new biomedical signal processing approach basedon the dual-tree complex wavelet transform is proposed. The dual-tree complex wavelettransform with translation invariance, anti-aliasing and directional selectivity, canovercome these shortcomings of traditional discrete wavelet transform. In the paper, thedual-tree complex wavelet transform is used in noise removing and feature extraction orfusion images of biomedical signals, including impedance differential signal, ECG,heard sound signal and medical image. The experimental results show that comparedwith the traditional discrete wavelet transform, the dual-tree complex wavelet transformreduces noise more thoroughly, retains and extracts boundary and texture characteristicsbetter. And the fusion image is more clear, and includes more information. So it can beused as a new method for biomedical signal processing.