This talk will introduce particle filtering algorithms designed for beat tracking in musical audio signals, i.e. to output a pulse which corresponds to the preferred (trained) human tapping tempo. The first stage of the process is to extract musically relevant change-points from the signal; the second half of the problem is to find a tempo profile which fits these and quantises them to "score locations". Particle filters are a stochastic estimation tool which are suitable for sequential estimation and we will present two models, one based upon the traditional Kalman filter and one which represents tempo evolution as Brownian motion. Results have been tested upon a large and varied database and are comparable with the state of the art.