Speech taken apart and rebuilt before a hall
On 11 September 1936, at the Harvard Tercentenary Conference in Cambridge, Homer Dudley of Bell Laboratories accompanied F. B. Jewett's address with a demonstration of a system that analyses speech and at once synthesises it anew. A voice spoken in a laboratory in New York was resolved by an analyser into pitch, resonance regions of frequency and loudness, and a synthesiser built a new signal from those control currents, which went by line to the hall in Cambridge.
Why it matters
This is the earliest demonstration in the atlas in which speech is not transmitted but taken apart into parameters and built again from electrical counterparts of the vocal cords and the mouth's cavities. An editorial assessment: a scheme whose control signals can be altered between analysis and synthesis lies beneath later speech coding and synthesis; the December article does not yet say so.
By Dudley's article, speech originating in the laboratory in New York was picked up by a moving-coil microphone; the analyser determined the three fundamental characteristics of speech sounds: pitch, resonant frequency regions and loudness. The control currents went to the synthesiser, which produced a new current to their specification, and a line carried it to the hall in Cambridge, where amplifiers turned it into sound. The synthesiser mirrors the human vocal system: batteries in place of air from the lungs; a buzzer-sounding relaxation oscillator in place of the vocal cords for voiced sounds; a resistance noise for the unvoiced sounds made at a constriction in the mouth; tuned electrical networks in place of the resonant chambers of the mouth. The currents from the analyser can be altered on their way: the pitch raised, lowered or held fixed, varied by a manually operated dial independently of the talker's own pitch, and the artificial voice can be used together with the original to produce a duet from a single voice. Dudley himself says the system is not perfect, though it produces easily intelligible speech with the usual variations of pitch and loudness. What the record does not state. The article gives no number of channels or bands, no delay and no intelligibility figure, and the word vocoder does not appear in it: numbers of that kind come only with the 1939 paper. It does not say who spoke during the demonstration itself: the photograph on its first page is captioned "C. W. Vadersen speaking into the microphone of the speech-synthesizing apparatus" and shows the apparatus in the laboratory, so it does not prove he was the talker in the hall. The hall is named two ways: "Sanders Hall" in the caption and on the diagram, "Sanders Theatre" in the text. The content of Jewett's address is not set out here: no account of it was read.