摘要 |
New dialogue is post-synchronised with guide track dialogue by using signal processing apparatus in which the analog guide track signal x1 (t) undergoes speech parameter measurement processing in a processor 43 to provide a speech parameter measurement processing in a processor 43 to provide a speech parameter vector A (kT). The new dialogue signal x2 (t min ) is processed to give waveform data which can be stored on disc 25 and a speech parameter vector B (jT) from a parameter extraction processor 42. The variables k and j are data frame numbers, and T is an analysis interval. Some parameters of the vector B are used in a process 48 to classify successive passages of the new dialogue signal into speech and silence, to produce classification data f (jT). The vectors A and B and the classification data are utilized in a time warp processor SBC2 to determine a timewarping function w (kT) giving the values of j in terms of the values of k associated with the corresponding speech features, and thereby, indicating the amount of expansion or compression of the waveform data of the new dialogue signal needed to align the time-dependent features of the new dialogue signal with the corresponding features of the guide track signal. Editing instructions are generated in signal editor computer SBC1 from the w (kT) data, feature classification data, pitch data p (jT) and the data stream x2 (nD) so that the editing of x2 (nD) can be carried out by the computer SBC1 in which periods of silence or speech are lengthened or shortened to give the desired alignment. The edited data x2 (nD) is converted to analog by a converter unit 29. and low pass filtered to provide an audio output signal to be recorded as the synchronised new dialogue. |