GB1387321A - Voice synthesizer - Google Patents

Abstract

1387321 Speech synthesiser R T GAGNON 15 Feb 1972 [1 March 1971] 6966/72 Heading H4R In a speech synthesiser an 8 bit input signal is fed into a register 76, two of the digits are used to control the frequency of a voiced excitation function generator 14 while the remaining digits are fed to a memory matrix 80 which converts the 6 bit signal into a particular combination of binary signals on its output leads. The output leads are connected to resistor adding networks 82a to 82j to provide a number of analogue control signals which control, in the synthesiser 18, the characteristics of the excitation signal, the frequency and amplitude of formants synthesised by the tunable filters 42, 44, 46, and the duration of a speech phoneme synthesised by the apparatus. The analogue control signals are generally applied to the synthesizer via low pass filters 84a to 84h, to make the transistion between phonemes smooth. As described, a resistive adder 82a controls an exponential function generator 86 to produce, when required, a step-function followed by an exponential decay which, when applied to analogue multiplier 30 connected to the noise generator 16, provides a modulation of the excitation function to generate a "plosive". A further adder 82d via a low pass filter 846 controls a chopper 92 which is fed with a signal from the voiced excitation function generator and also controls multiplier 30 to provide a pitch frequency modulation of the noise generator during production of a voiced fricative, such as "b" "z", &c. Resistive adders 82f, 82e, 82d, control the frequencies of respective filters 42, 44, and 46, via respective filters, 84f, 84e, 84d to provide formant frequency emphasis. Analogue multipliers 60, 64, and 27 provide control of the relative amplitudes of the formants and a background excitation function signal, according to the outputs of further adder networks, and an adder network 82j, via timing control 98, controls the duration of the phoneme corresponding to the input signal fed to register 76. A fixed resonant filter 69 at the output of the synthesizer provides a fourth formant corresponding to the natural resonant frequency of the nasal cavity in a speaker. Circuits are described for the various parts of the synthesizer.