This paper addresses the restoration of audio signals proceeding from old recordings, and focuses on long-pulse removal. We propose a new two-stage method to estimate the waveform of each long pulse from the observed noisy signal. First, an initial estimate for the pulse shape is obtained via a non-linear filtering scheme called two-pass splitting-window (TPSW) filtering. Then, this estimate is further smoothed through a piecewise polynomial fitting. The degree of smoothness of the estimate can be controlled by adjusting either the TPSW parameters or the length of the segments to be fitted. The proposed method has low computational complexity, it is not constrained by the assumption of shape similarity among pulse waveforms, and can be successfully applied for removing overlapping pulses.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.