FNELSON@OBERLIN.BITNET (02/10/90)
LPC is a software synthesis technique that can produce very realistic results by analysing sampled sounds and separating the spectrum from the pitch and sound source. Speaking voices can be made to sing., acoustic instruments can whisper. One of the greatest examples is "lPerfect Days" by Andy Moorer where a poet and a flutists are recorded and cross-synthesized in ways that are impossible to describe. Imagiune a vocoder with much wider timbral characteristics. Some of Lansky's pieces are out on CD now (don't know the numbers) but look in the Schwann catalog). "idleIdle Chatter" is his biggest hit but allm all of his pieces are woirth listening tooto. AlonPaul Lansky and Charles Dodg e usareboth use LPC.
adamson@ITD.NRL.NAVY.MIL (Brian Adamson) (02/10/90)
An example of what can be done with LPC (and has been done here) Taking a monotone spoken sentence and making it sound like a barber shop quartet.. Having Fred speak into the processor but having the speech come out on the other end in Barney' voice. It is also the basis for low-data rate speech. . 2400 bits per second 800 & 600 bps speech on the horizon (real-time speech, by the way) Brian Adamson
YEIDEL@WSUVM1.BITNET (Joshua Yeidel) (02/10/90)
The following note appeared on the NEXT-L list. I don't understand it in detail, but some of you might. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ------------------------------------------------------------------------ From: pmy@jeeves.acc.Virginia.EDU (Pete Yadlowsky) Subject: Announcing... Date: 8 Feb 90 00:36:28 GMT Organization: University of Virginia, Charlottesville ....the release of LPC, a NeXT adaptation of Paul Lansky's linear predictive analysis and pitch tracking programs. LPC reads a NeXT or CSound-NeXT format soundfile and produces a data file which codes the sound in terms of its time-varying formant characteristics and the RMS levels of a residual excitation signal. To resynthesize the audio signal, this LPC-derived dynamic filter is excited with a spectrally rich driver, such as white noise or a pulse train, though many interesting effects can be produced with other types of signals. Also, there is a pitch tracking utility which produces data that may be merged with the LP analysis data. This pitch information may then be used to control the pitch of the driving signal during resynthesis. CSound (Barry Vercoe's software synthesis system) provides signal modification units which read LPC analysis files and resonate specified signals through the LP-encoded filters. CSound has also been ported to the NeXT. The next release of CSound-NeXT (sigh) will have hooks for launching and communicating with LPC-NeXT, thus adding another application to the CSound-NeXT integrated tool set. Due out shortly. CSound-NeXT and LPC-NeXT are both NeXTStep applications, meaning that their front ends are built on the NeXT graphic user interface for clear, intuitive interaction. For a detailed explanation of the workings of linear predictive analysis, try "Computer Music" by Dodge and Jerse. Hal Chamberlin's "Musical Applications of Microprocessors" also has a few paragraphs on this. ftp: uvaarpa.acc.virginia.edu:/pub/next/Apps/LPC.tar.Z uvaarpa.acc.virginia.edu:/pub/next/source/LPC-source.tar.Z Piece-wise uuencoded mailings can be arranged for those who do not have ftp access. Pete Yadlowsky Academic Computing Center University of Virginia pmy@Virginia.EDU Peter M. Yadlowsky | "Pay no attention to that man Academic Computing Center | behind the curtain!" University of Virginia | pmy@Virginia.EDU |