Skip to Main Content
In this paper, we propose a new algorithm for automatic transcription of music signals. This algorithm consists of two main stages: the first stage eliminates the harmonics of music signal and only passes the fundamental part. This will extremely simplify the modeling of music signal. The second stage estimates and tracks the fundamental frequency of music signal by means of an Extended Kalman Filter (EKF) frequency tracker. In this way, we will identify each note and its time duration. Our approach does not face any windowing problem and it is suitable for on-line music transcription.