Page 44 - Special Issue

Early Talking Automata
A
C
B
D
Figure 3. Kempelen’s speaking machine. A and B: photographs of the mechanism exhibited in the Deutsches Museum in Munich, Germany, thought to be a reconstruction of the original. C and D: corresponding illustrations from Kempelen (1791). D: details shown (see text for discussion) include the bellows (X), the box containing the vibrating reed (A), and the mouth tube (C); the other pipes (m and n) and levers (r and sch) are extra modifications needed for nasals, fricatives, and trills. Photographs reproduced from the Deutsches Museum, Munich, Germany, with permission. Copyright Deutsches Museum, München, Archiv, BN37402 and BN37404.
The same elements persisted from previous attempts. A pair of bellows and a tube mimicked the lungs and trachea and a set of valves directed airflow into a collection of boxes. The entry of each box was covered with a leather diaphragm over an elliptical hole representing the glottis, while a parchment reed covering the hole vibrated like the vocal folds; by moving a metal tongue over the reed, the tone could be adjusted. The different actuators shaping the inside of each box simulated movements of the vocal tract with levers and shutters pulled by cords. Vowels and diphthongs were produced by connect- ing particular boxes in sequence. Stops were produced by rapidly opening and closing shutters over the ends of boxes. Fricatives were synthesized by silencing the reed and blowing air into the boxes. Trills were produced by a special vibrating reed. Syllables were produced by sequences of movements built into the actuators for each box. Unlike previous mecha- nisms, the boxes seem to have modeled dynamic articulations approximating specific consonant-vowel combinations.
144 | Acoustics Today | Suprminmge2r0201,9Special Issue Reprinted from volume 15, issue 2
Most importantly, following Engramelle and Vaucanson, a pinned cylinder was turned to activate the different articula- tions in sequence, and the position of the pins could be altered to program whole sentences. Examples of utterances given in newspaper reports from the period include single vowels, a, e, and o; diphthongs, oa; syllables, pe, la, le, fe, fai, ra, and ro; and the following extended series of sentences mimicking a conver- sation between the two heads that was presented before Louis XVI at Versailles in September 1783 (see Figure 4 for transla- tion). The 1st Head begins, “Le Roi donne la paix à l’Europe”; the 2nd Head replies, “La paix couronne le Roi de gloire”; the
1st Head responds to the 2nd head, “Et la paix fait le bonheur des peuples”; and then addresses the King, “O Roi adorable, père de vos peuples, leur bonheur fait voir à l’Europe la gloire de votre trône.” Public reaction was divided. Many thought the display was sensational as proof that speech could be synthe- sized mechanically, whereas others complained that they could barely understand what was being said.