A new robust signal analysis method against non-stationary burst noise suitable for IoT

In this paper, we propose a period estimation method suitable for analyzing time-varying signals including burst noise. Frequency analysis method such as Fast Fourier Transform (FFT) are known to have low resolution in low-frequency bands, making it difficult to analyze low-frequency signals such as vibration of large structures or factory equipment. To solve this problem, we proposed accumulation for real-time serial-to-parallel converter (ARS). However, it was found that both FFT and ARS cannot accurately estimate the period of a signals containing burst noise. Furthermore, it is suggested that masking processing applied to signals containing burst noise can reduce its impact, but this reduces the accuracy of period estimation. This phenomenon makes it impossible to use period estimation methods in burst noise environments. we propose a new method to improve the burst noise tolerance of ARS. Furthermore, we confirm that the proposed method can analyze time-varying signals with burst noise.