508 the bell system technical journal, march 1954 



Harris, C. M.^ 



Speech Synthesizer, Acoust. Soc. Am., J., 25, pp. 970-975, Sept., 

 1953. 



"Standardized speech" constructed from building blocks called speech 

 modules has been described; it was sjmthesized by piecmg together bits of 

 magnetic tape containing recorded speech sounds. An electromagnetic 

 device, a "speech module synthesizer," is described here which performs 

 the synthesis automatically. When buttons on a kej'board are pressed, a 

 sequence of corresponding speech modules are automatically recorded on 

 tape exactly in tandem. The modules are selected from a group "stored" 

 on a rotating magnetic drum. The pressing of a button causes an electrical 

 signal corresponding to a module to be reproduced — the electrical switch- 

 ing is so arranged that onlj' one complete module is reproduced for a single 

 button-pressing. This electrical signal is amplified, biased, and then fed 

 into a constantly rotating head which makes contact with stationary mag- 

 netic tape and records the signal on it. A 10-kc signal superposed on each 

 stored speech module controls an electromagnetic clutch which (a) measures 

 the length of the recording accurately', and (b) advances the tape at the 

 completion of the recording by the correct amount so that the next record- 

 ing forms a connected sequence with it. The same module may be used any 

 number of times and in combination with different stored modules, thereb}- 

 introducing wider experimental control in standardized speech studies. 

 The principle of this type of device could be applied to other classes of 

 problems involving communication of information, as the conversion into 

 speech of typing or of electronicallj'-red printed matter. 



Harris, C. M.^ 



Study of the Building Blocks in Speech, Acoust. Soc. Am., J., 25, 

 pp. 962-969, Sept., 1953. 



Identification of the information-bearing elements of speech is important 

 in applj'ing recent tl linking on information theory to speech communica- 

 tion. One way to study this pi'oblem is to select groups of building blocks 

 and use them to form standardized speech which then ma}- be evaluated; 

 a method having the advantage of simplicit}' is described. Individual re- 

 cordings of the building lilocks were made on magnetic tape and then var- 

 ious pieces of tape were joined together to form words. Experiments indica- 

 ted that speech based upon one building block for each vowel and consonant 

 not only sounds unnatural but is mostly unintelligible because the influences 

 on vowel and consonants aie niissing which ordinarih* occur between ad- 

 jacent speech sounds. To synthesize speech with reasonable naturalness, 

 the influence factor should be included. Here these influences can be ap- 

 ])roxi mated l)y em])loying moi-e than one l)uilding block to represent each 

 linguistic element and by selecting these blocks properly, taking into account 

 the spectral characteristics of adjacent sounds so as to approximate the 



^ Bell Telephone Laboratories, Inc. 



