| 摘要 |
1,081,720. Speech recognition. STANDARD TELEPHONES & CABLES Ltd. Nov. 5, 1965, No. 46984/65. Addition to 1,012,765. Heading G4R. Speech recognition apparatus includes means for generating a non-linear pulse train time scale when a reversal of polarity is detected in the speech waveform and means for counting the number of timing pulses between said reversal and the next reversal, the initial time constant of the time scale being controlled by a voltage proportional to the fundamental frequency of the speech waveform. The pulses of the pulse train get progressively further apart. A negative input pulse of amplitude proportional to the fundamental frequency of the speech waveform charges a first capacitor, which then starts to discharge. At the same time, a second capacitor starts to charge negatively until the voltage across it is sufficiently negative to cause a first transistor to conduct. The first transistor conducts progressively more strongly due to positive feedback via a second transistor until it is saturated when it discharges the second capacitor through grid drain until the second capacitor cannot maintain the first transistor conducting when the latter cuts-off. While conducting, the first transistor produced the first clock pulse. The second capacitor is now charged from the first but not so fast as before due to loss of charge from the latter in the meantime. The second capacitor in due course causes the first transistor to conduct again to produce the second clock pulse, and so on. The series of clock pulses terminates when a further negative input pulse arrives to produce a new clock pulse series. The further input pulse may be preceded by a positive resetting pulse applied to the second capacitor. |
