Extraction of transient signals using adaptive line enhancer

Abstract

This paper proposes a method of extracting transient signals in the presence of colored noise using an adaptive line enhancer. Transient signals are non-stationary signals with a shorter duration than the observation interval. The problem of transient signal extraction arises in many fields of application - for instance, in mechanical failure diagnostics and in underwater acoustic warfare - where the transient signals usually contain a wealth of useful object feature information. But it is very difficult to extract transient signals with only a few oscillation cycles in a strongly colored noise background. The traditional method of improving the signal to noise ratio is not very effective. Some people try to use the Hilbert-Huang Transform method, but the method itself has many problems, such as the lack of rigorous physical meaning, and probable existence of serious waveform distortion. We have trialled the use an adaptive line enhancer (ALE) to extract the transient signals with encouraging results. Here an adaptive line enhancer is a degenerate form of the adaptive noise canceller in that its reference signal consists of a delayed version of the input signal. According to the change of the input signal, the adaptive filter coefficients are updated constantly by using a recursive structure. An adaptive line enhancer can separate the input signal into a strong correlation component as the filtered output and produce a weak correlation component as an estimation error. Transient signals will emerge in the error component when the adaptive algorithm convergence speed is slow enough. If the interference noise is colored noise sharing the larger proportion in the observed signal, the above methods can achieve high processing gain with little waveform distortion even if the frequency of the strong interfering signal and transient signals are the same. Here, we will describe the complete process that separates transient signals from the colored noise by using ALE. Finally, computer simulation results are presented. The simulation results indicate that: when the step parameter value is small enough, the transient signal can be extracted perfectly from the strong noise background; but with the increase of the step parameter value, the waveform distortion will increase seriously.