Open main menu

Page:Popular Science Monthly Volume 89.djvu/82

This page needs to be proofread.


��Popular Science Monthly


01 5«y __OCl Sec


��well as on experimentation, back of which were the resources of a great t\"pe- writer company.

The line of reasoning in- volved in designing the machine, though somewhat intricate, is exceedingly in- teresting. Getting any machine at all to respond to such an uncertain and vari- able director as the human voice, is a task beset with difficulties.

Speech Had First to Be Studied

In a recent paper which he read before the American Institute of Electrical En- gineers, the inventor dis- cussed researches lately completed into the true nature of speech, these having a great deal to do with the practical workings of the eventual machine. It was discovered that all speech can be represented by a sort of natural alphabet of sound patterns, which, no matter what the voice may be, always have the same shape. When a man. for instance, pronounces a given word he molds air waves in precisely the same way as does a woman. So far as sounds go, a Choctaw Indian is as well provided as a Har\-ard graduate; the onl\- difference is that the sounds are grouped differently. This is a fundamental law. The mechanism of sf)eech is the same in all of us. Heretofore physicists and workers with speech and sound have been troubled by the fact that they had nothing definite to work with. The consonant letters, when one person spoke them, would appear to have much the same wave shape as vowels enunciated by another speaker. In fact, even consonants and vowels produced by the same person sometimes seemed to have these indeterminate shapes when the scientists squinted at them through their sound-wave record- ing machines. Hence the task of ever getting spoken sounds analyzed and classified for study seemed hopeless. Until these letter-sounds were analvzed



��After all, what are spoken words but telegraph signals sent through the air, collected by the ear, and interpreted by your brain? Consider spoken words as sound signals and the voice-operated type- writer becomes possible

Alphabet of natural leXter- patierns obtained icith the ap- paratus shown at top of page 66. Xotg that the natural alphabet is not unlike that now used in submarine-cable telegraphy, though of course the two alpha- bets have no connection, theoret- ical or otherwise. The machine shcrwn on page 67 spells out words in this natural alphabet.

This phonetic writing may some day be used in offices as a sort of short-hand system, the dictator talking into a machine similar in principle to that shown on page 67, and the stenographer afterward reading the wavy line from the roll of paper as easily as she would her own notes. The machine with its present design is entirely in laboratory form — interesting hcru:- ever, for the novelty of the idea cm -j:hich it is based, and because it comes closest to tracing the true wave-form of speech of any machine yet devised.

��and classified so that somebody could reason out the real underlying law they obeyed, it was obviously impossible to go go far toward a \oice-operated type- writer. One cannot simply say "Write!" to an inanimate collection of levers and expect them to respond.

Why Whispers Were Studied

The instruments which were used in determining the real nature of speech sounds are shown on Page 66. With this apparatus Mr. Flowers dealt only in whispers. Why? Because whispering is the most elemental way one can convey speech. When you whisper you make no use of the vocal cords or other com- plicated mouth and throat mechanisms. It may be said in passing that one of the principal reasons previous workers with speech sounds failed to get at true sound-wave shapes was that over-tones (extra tones that cause a given voice to have its peculiar and distinctive nature) caused the shape of the main tone or fundamental to be obliterated. Res- onant or echoing tones arising in the

�� �