BOLT BERANEK AND NEWMAN, INC. (BBN)

BBN Laboratories, Inc.
Cambridge, MA


------------------------------------------------------------- Top
HISTORY OF ORGANIZATION

  need history synopsis


------------------------------------------------------------- Top
PROJECT: ANALYSIS/SYNTHESIS VOCODER USING LPC DIPHONES (19  -19   )


Speech synthesizer portion of a vocoder that operates from an
input string of desired phoneme name, duration, and pitch value.
The synthesizer first translates the phoneme string into a
corresponding diphone sequence, then assembles the synthetic
speech from a stored library of LPC-encoded diphone segments. The
diphone segments are time-warped to meet the duration
requirement, abutted, and smoothed to reduce the perceptual
effect of mismatches in the LPC parameter values at the points of
juncture. The goal was to attain a vocoder data rate of less than
100 bits per second.


1977 Makhoul, J., C. Cook, R. Schwartz, and D. Klatt, "A
feasibility study of very low rate speech compression systems,"
Report No. 3508, Bolt Beranek and Newman Inc., Cambridge, MA, Feb
1977.  (****  need copy  ****)


1979 Schwartz, R., J. Klovstad, J. Makhoul, D. Klatt, and V.
     Zue,"Diphone synthesis for phonetic vocoding", ICASSP-79,
     891-894, 1979.


Examples of diphone synthesis (output of the vocoder with phoneme
errors corrected by hand) and automatic operation as a vocoder.
The number of phoneme errors in recognition was deemed too large
for practical use. (personal communication from R.M. Schwartz to
H.D. Maxey, 5/8/85, in SSSHP USA Bolt Beranek and Newman file)

     SSSHP 116 Tape: BBN Phonetic Vocoder, from R. Schwartz, 5/8/85
          ("Rainbow Passage" by: Phonetic synthesis, phonetic
          vocoder, original speech)
          5" reel, 7 1/2 ips, good quality


------------------------------------------------------------- Top
PROJECT: TESTING OF TEXT-TO-SPEECH SYSTEMS (19   - 19   )


1986 Huggins, A.W.F., R.A. Houde, and E. Colwell, "Vidvox human
     factors investigation," Bolt Beranek and Newman Inc., Report
     No. 6187, Cambridge, MA, p. 15 (1986).  Test of 1600 low-
     frequency polysyllabic words. Speech Plus Calltext System:
     8.3% phonemic mispronunciations. DECtalk Vers. 1: 12.9%.  (K)

                    (***** need copy *****)

------------------------------------------------------------- Top
BIOGRAPHIES


E. COLWELL

C. COOK

ROBERT A. HOUDE

A.W.F. HUGGINS

DENNIS H. KLATT  (see SSSHP USA MIT file)

J. KLOVSTAD

J. MAKHOUL

RICHARD M. SCHWARTZ

VICTOR ZUE

1968 B.S. in Elect. Engr., Univ. of Florida, Gainesville
1969 M.S.E. in Elect. Engr., Univ. of Florida, Gainesville


------------------------------------------------------------- Top
CONTRIBUTIONS FROM:

Mr. Richard M. Schwartz
BBN Laboratories Incorporated
10 Moulton Street
Cambridge, MA 02238
 
 
REVIEW BY: