September 2, 



1922] 



NA TURE 



An Electrical Analogue of the Vocal Organs. 



In connexion with correspondence which recently 

 has appeared in the columns of Nature relating to 

 the physical characteristics of vowel-sounds, the 

 following account may be of interest of an apparatus 

 believed to be novel, which is conveniently capable 

 of the artificial production of many speech-sounds. 

 It is well known that Helmholtz succeeded in imitat- 

 ing vowels by combinations of tuning forks, and 

 Miller by combinations of organ pipes. Others, 

 notably Scripture, have constructed apparatus wherein 

 the transient oscillations of air in resonant cavities 

 were excited by series of puffs of air, in close physical 

 imitation of the action of the human vocal organs. 

 It seems hitherto to have been overlooked that a 

 functional copy of the vocal organs can be devised 

 which depends upon the production of audio-frequency 

 oscillations in electrical circuits. 



A schematic diagram of such an apparatus is given 

 in Fig. 1. Periodic interruptions of the electric 

 current, produced by a buzzer or a motor-driven 

 circuit interrupter, corresponded to the periodic inter- 



2t VoLT5 



ruptions of the air current in the human throat by 

 the vocal cords. The intermittent electric current 

 thus produced excited the natural damped oscillations 

 of the resonant circuits, / and 2. This was confirmed 

 by observation with an oscillograph. In like manner, 

 puffs of air from the vocal cords excite the natural 

 damped oscillations of the air in the mouth cavities. 

 The work of numerous investigators has indicated 

 that the air in the mouth cavities possesses, as a 

 rule, only one or two important modes of vibration. 

 The oscillations of the electric current were trans- 

 formed to sound-vibrations in the air by a loosely 

 coupled telephone receiver. The distortion intro- 

 duced by this telephone receiver appeared to be of 

 little importance. 



Appropriate adjustments of the resonant circuits 

 / and 2 were observed to result in the production of 

 all the various vowels and semi-vowels 111 turn. 

 Alteration of the frequency or damping of either 

 resonant circuit was observed to result in alteration 

 of the vowel produced. The frequency of interrup- 

 tion, which was the group frequency of the recurrent 

 damped oscillations, was observed to determine the 

 pitch of the vowel ; but it did not determine what 

 vowel was produced. Similarly, in the case of the 

 human voice the frequency of vibration of the vocal 



NO. 2757, VOL. I IO] 



cords is known to determine the voice-pitch, while 

 the adjustment of the mouth cavities is known to 

 characterise the vowel. 



The vowels and semi-vowels produced by the 

 " electrical voice " with regular interruptions, in the 

 manner just described, were equivalent to intoned or 

 sung vowels, or, if the frequency of interruption was 

 made to vary appropriately, to spoken vowels. It 

 was found possible to produce the whispered vowels 

 with interruptions that were non-periodic, which is 

 in accordance with the idea that, in the human voice, 

 whispered speech is due to irregular frictional modula- 

 tion of the exhaled air. Whether the vowel was 

 whispered, sung, or spoken depended upon the manner 

 of making the interruptions ; while what particular 

 vowel was produced depended upon the adjustment 

 of the resonant circuits. 



Diphthongs were produced by altering the circuit 

 adjustments rapidly so as to shift from the initial to 

 the final vowel-sound of the diphthong pair. Some 

 of the fricative (hissing) consonants were approxi- 

 mated with irregular interruptions, provided the 

 resonant circuits were set at somewhat higher fre- 

 quencies than for the vowels and semi-vowels. None of 

 the explosive consonants were satisfactorily imitated. 

 It is believed that lack of success with the explo- 

 sives was due to obvious difficulties of manipulation. 



There was much room for improvement with respect 

 to the naturalness of the " electrical voice." It was 

 too monotonous, as was to have been anticipated. 

 Contrary to expectation, alteration in the wave-form 

 of the exciting current did not materially change the 

 tone, provided the wave-form was sufficiently far from 

 sinusoidal. The intoned vowels, semi-vowels, and 

 diphthongs produced by the-" electrical voice " were 

 sufficiently natural to be recognised in at least fifty 

 per cent, of the trials by eight or ten different ob- 

 servers. When arrangements, not indicated in Fig. 1, 

 were made to give the appropriate circuit adjustments 

 in rapid succession, simple words like " mama," 

 " Anna," " wow-wow," yi-yi," were fairly well imi- 

 tated. The whispered vowels and fricative con- 

 sonants were not imitated so well, because (it is 

 thought) of the lack of complete irregularity in the 

 circuit interruptions. In human speech the pitch 

 and vowel quality and intensity are constantly 

 changing in a way difficult to imitate with the crude 

 apparatus of Fig. 1. Probably, also, in the human 

 voice additional weak transient oscillations are ex- 

 cited, due to minor modes of vibration of the air in 

 the cavities of the head, which determine the in- 

 dividuality of the voice without greatly altering the 

 speech-sounds. 



Thus these experimental results are sufficient to 

 give a general qualitative description of each of the 

 following four classes of speech -sounds : vowels, 

 semi-vowels, diphthongs, and fricative consonants (but 

 not explosives), whether sung, spoken, or whispered ; 

 and these results also make possible quantitative 

 specification of the characteristics of the various 

 speech-sounds themselves. Numerical values of the 

 frequencies and dampings which appear to charac- 

 terise the various speech-sounds have been calculated 

 from the electrical constants possessed by the resonant 

 circuits of the apparatus when adjusted to produce 

 them. 



The analytical expression for a single transient due to 

 one resonant circuit when loosely coupled is, of course, 

 ['Instantaneous displacement = Ae~ at sin 2irft, 



where e is the base of natural logarithms, a the 

 damping constant, f the frequency, and A the ampli- 

 tude. The displacement in the air-vibration is taken 

 as proportional to the instantaneous current. Also, 

 / = i/27rsLC nearly, and a=R/2L. The capacity C 



