Xref: utzoo unix-pc.general:1491 comp.sys.att:4433 Path: utzoo!attcan!uunet!ncrlnk!ncrcae!hubcap!gatech!ncar!boulder!pikes!udenva!isis!scicom!zebra!vern From: vern@zebra.UUCP (Vernon C. Hoxie) Newsgroups: unix-pc.general,comp.sys.att Subject: Voice Power Keywords: answer machines, speech recognition, text to speech Message-ID: <119@zebra.UUCP> Date: 2 Oct 88 05:08:57 GMT Organization: Alpha Science Computer Networks, Denver, Colo. Lines: 50 The principal application of the UNIX-PC's VOICE POWER board has been as a telephone answering machine. It seems to have been developed to be used in groups of seven cards using the expansion chassis. The application usually is to ask questions of the caller and have them respond with touch-tone signals for more information. Thus a caller can step through a query-response tree to obtain information rather than have live people answering the phone at places such as newspapers, brokerage houses and the like. It can also be used as a poor mans AUDIX by recording the callers voice just as a tape recorder does. With the digital recording, each message is a separate file and may be saved or deleted as desired. However, big disks are a must. The thing eats up memory at the rate of 100,000 bytes per minute. This is using the 16 kb sub-band mode. I haven't tried the optional 24 kb or 64 kb modes yet. Another application recently rumored was to sell it with the System-75 for moderate sized customers. They could have it answer the phone with a canned response. Knowledgable inward callers can respond by dialing an inside extension number. This reduces the cost of paying local operating companies for direct inward dialing. Divestiture has changed AT&T's attitude hasn't it? There are a number of playthings developers might have fun with such as the 'voice editor' and the 'text to speech' programs. The voice editor allows you record to disk then cut and paste the sounds just as you move text with an ascii editor. The test to speech program reads an ascii file in a synthesized voice. The voice has male qualities and cannot be changed. I have the suspicion that this is built into the firmware attached to the board. He does pretty well on regular words but reverts to spelling if he cannot figure out a correct pronunciation. A cut line becomes boring when he reads each '-' as 'dash, dash'. There is also the capability to initiate outward calls with its own tone generator from prerecorded tones. This infers that another application is for programmed telephone soliciting. I believe that such a practice has been outlawed in Colorado. We get enough live people peddling junk by phone anyhow. The voice files are standard unix files and are available for processing by user developed programs such as sound filter or analyzed by fourier transforms etc. Similar cards are also vended by AT&T for there 6312 and 6386 machines. I would be interested in learning of any applications or software anyone has developed they would care to share. My intent was to try play with voice to text developments. So far I have involved with other projects such as working for money. How droll! >>> Every expert was once a novice. <<< Vernon C. Hoxie scicom!zebra!vern 3975 W. 29th Ave. voice: 303-477-1780 Denver, Colo., 80212 UUCP: 303-455-2670 Daddy, when is it going to be my turn to be an expert? Huh, daddy? Huh??