Timbre morphing and interpolation based on a sinusoidal model
Naotoshi Osaka · The Journal of the Acoustical Society of America · 1998
There have been several attempts to synthesize the timbre in between two sounds. Several synthesis methods are reported to work well, but are not automatic throughout the whole process or are not precisely described if one wishes to trace the algorithm. An automatic algorithm of synthesizing morphed sounds and interpolated timbre has been reported. Two original sounds are analyzed using the M&Q algorithm. Correspondent partials of each sound are found using Dynamic Programming. Two synthesis methods are introduced using a sinusoidal model. One is to find correspondent partials within correspondent frames of two sounds. Then an interpolated partial of a new frame is calculated. Peak-to-peak match of sinusoidal parameters is done for a new frame and synthesized sounds are acquired. The other method is the one which changes the procedure sequence of the first method. Correspondence is considered for sequent partials and not a framewise signal. These two methods are compared with another simple synthesis method, that is, wave-to-wave interpolation of one pitch period. Total performance of sounds are discussed using subjective test results.