The analysis of vibrations from rotating machines gives information about their faults. From the signal processing perspective a significant problem is the detection of weak signals embedded in strong noise. Stochastic resonance (SR) is a mechanism where noise is not suppressed but exploited to trigger the synchronization of a non-linear system and in its one-dimensional form has been recently applied to vibration analysis. This paper focuses on the use of SR in a two-dimensional system of gradient type for detection of weak signals submerged in Gaussian noise. Comparing the traditional one-dimensional system and the two-dimensional used here, this paper shows that the latter can offer a more sensitive means of detection. An alternative metric is proposed to assess the output signal quality, requiring no a priori knowledge of the signal to be detected, and it is shown to offer similar results to the more conventional signal-to-noise ratio.
|Title of host publication||Proceedings of the 23rd European Signal Processing Conference, 2015|
|Publication status||Published - 2015|
|Event||Signal Processing Conference (EUSIPCO), 2015 - Nice, Italy|
Duration: 31 Aug 2015 → 4 Sep 2015
|Conference||Signal Processing Conference (EUSIPCO), 2015|
|Period||31/08/15 → 4/09/15|
Barbini, L., Cole, M. O. T., Hillis, A. J., & Du Bois, J. L. (2015). Weak signal detection based on two dimensional stochastic resonance. In Proceedings of the 23rd European Signal Processing Conference, 2015 IEEE. https://doi.org/10.1109/EUSIPCO.2015.7362764