Tracking pitch period using particle filters
Geliang Zhang, Simon Godsill · 2013
Pitch tracking has been used in many speech processing applications. Most present time domain techniques in pitch estimation mainly use autocorrelation methods and the average magnitude difference functions. This paper aims to track pitch period of speech using the particle filter approach. A simple model has been proposed to capture the pitch period variations of noisy speech during voiced periods. Performance of the proposed method is compared with standard pitch detection algorithms. Simulation results show that the proposed method can track the pitch period even if strong noise exists. It suggests that the particle filter approach could be an alternative way to address the pitch tracking problem.