A Flexible Analysis-Synthesis Method for Transients
Year: 2000
Type: conference-paper
Abstract: We propose a new transient model which naturally extends the sinusoidal model to nonstationary phenomena. An integrated online-o#ine segmentation approach is developed, followed by an explicitly time-variant sinusoidal model. 1 Introduction The sinusoidal model is a convenient soundrepresentation in analysis-synthesis frameworks. The signal is modeled as a sum of sinusoids, with timevariant amplitudes and frequencies, plus a residual r(n), or unmodeled part: y(n) = 2p # k=1 A k,n cos( # m<n # k,m + # k,n ) + r n Such a parametric, time-frequency representation is ideal for analysis-synthesis goals: the parametric nature of the model allows meaningful specification of transformations on sound material, while the "timefrequency " nature facilitates intuition in developing musically viable transformations. Classical sinusoidal models; e.g. [MQ86] use STFT as front end. Implicit is the assumption of local stationarity. Estimation biases occur to the degree local stationarity i...
Authors: Harvey Thornburg, Fabien Gouyon
Institution: Stanford University
Language: English
Topic: Music and Audio Processing
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1850188872
