Particle Filtering Approach to Adaptive Time-Delay Estimation

A particle filter algorithm is developed for the problem of online subsample time-delay estimation between noisy signals received at two spatially separated sensors. The delay is modeled as an adaptive FIR filter whose coefficients are determined by the tracker´s particles, and updated on a sample-by-sample basis. Efficient tracking of the delay parameter over time is ensured with the derivation of a global system model integrating the target dynamics for both near-field and far-field operation. Experimental simulations are carried out to assess the algorithm´s convergence and tracking performance, and demonstrate that the proposed method is able to efficiently track time delays with stationary signals as well as speech