Detection of signal jumps in noisy signals using an adaptive stochastic gradient method

This paper introduces a novel technique for the detection of jumps in noisy signals using an adaptive stochastic gradient method. This method is an improved and extended version of the conventional stochastic gradient (SG) operator, which utilizes a FIR Weiner filter. The novel contributions in this paper are two fold; namely, (i) introduction of a robust method for estimating signal autocorrelation coefficients, signal-to-noise ratio and noise variance, and (ii) adaptation of the Weiner filter coefficients in a block-by-block fashion. Improved performance of the proposed method over conventional SG and filtered derivative techniques is demonstrated using synthetic data.