Detection and tracking of multi-periodic signals

Periodic signal analysis is an important tool in signal processing, there are many phenomena that exhibit a periodic nature. A number of analysis techniques aimed at estimating the periodicities from sensor data already exist but most use stationary harmonically related Fourier components as the basis. The performance can be seriously degraded when there are multiple signals present and/or a period is time-varying. In this paper a novel time-domain tracking method is proposed. This is based on a modified Incremental Multi-Parameter (IMP) algorithm [1] that is able to detect and track several periodic components in a single time series. The method exploits pseudo-integration, a novel method aimed at reducing tracking lag. Two forms of the algorithm are discussed: a) block mode, using an iterative approach on a batch of sampled data and b) recursive mode for updating parameters in a real-time practical situation.