Noisy signal parameter identification using Particle Swarm Optimization

Abstract

This work presents an approach to use Particle Swarm Optimization (PSO) to identify the amplitude, frequency and phase parameters of a sinusoidal signal corrupted with additive Gaussian noise using a discrete sample of it. We encode signal parameters in the particles and we evaluate its goodness by computing the root mean square (RMS) error of the difference between a discrete signal synthesized using the particle configuration and the input signal sequence. We have validated our approach by using a set of test signals presenting variations on their parameters and in the Signal to Noise Ratio (SNR) resulting from the signal corruption. The PSO was tuned by using a reference signal in order to choose a suitable configuration for the PSO parameters. The approach has shown to perform successfully with signals exhibiting a SNR as low as 16.99 dB with an RMS error of 3 %.