A unified approach to real time audio-to-score and audio-to-audio alignment using sequential Montecarlo inference techniques