Remaking Speech

Speech has been remade automatically from a buzzer-like tone and a hiss-like noise corresponding to the cord-tone and the breath-tone of normal speech. Control of pitch and spectrum obtained from a talker's speech are applied to make the synthetic speech copy the original speech sufficiently for good intelligibility although the currents used in such controls contain only low syllabic frequencies of the order of 10 cycles per second as contrasted with frequencies of 100 to 3000 cycles in the remade speech. The isolation of these speech-defining signals of pitch and spectrum makes it possible to reconstruct the speech to a wide variety of specifications. Striking demonstrations upon altering the pitch of the remade speech stress the contribution of the pitch to the emotional content of speech. Similarly the spectrum is shown to contribute most of the intelligibility to the speech.