Sound Source Localization Method of Substation Based on Fourth-order Cumulant

The performance of abnormal sound source localization technology for substation equipment based on acoustic sensor array is not only related to the processing algorithm, but also depends on the array model of the sensor. Increasing the number of sensors to improve the estimation performance will often lead to an increase in the cost. This paper proposed a non-redundant fourth-order cumulant array virtual expansion method based on the uniform planar array (UPA), which could improve the performance of the array without changing the hardware design. Compared with the low-order cumulant, the high-order cumulant increases the number of sensors through the virtual expansion of the array, and improves the localization estimation performance and the number of distinguishable sources. In addition, since the high-order cumulant only contains the information of non-Gaussian components, the expanded array has better anti-Gaussian noise performance. Simulation and field test results showed that the method proposed in this paper could achieve more localization estimations of sound source signals beyond the array elements of real sensors. The localization accuracy was significantly improved when the signal-to-noise ratio (SNR) was low, and the resolution of the localization method was improved.