SU1408450A1 - Method and apparatus for synthesis of speech signals - Google Patents

Description

(21) 4000887 / 24-10, 4000895 / 24-10

(22) 01/03/86

(46) 07.07.88. Bul Number 25

(71) Kiev Institute of Automation. XXV Congress of the CPSU

(72) G.F. Zakharchenko and I.V. Korkov

(53) 534.85 (088.8)

(56) Tadashi Kitamora, Satoshi Imai, Chieko Foroichi and Takao Kobayashi. Speech analysis and synthesis of speech usmg melcepstrom. Denshi Tsuchin Gakkaifembunshi, vol. 68-A, Sept. 1985, pp. 957-964.

Patent GDR ff 218970, CL. G 10 L 9/18, published 1985.

(54) METHOD OF SYNTHESIS OF SPEECH SIGNALS AND DEVICE FOR ITS OCyfflECTlMff W

(57) The invention relates to speech informatics. The purpose of the invention is to improve the dynamic properties of reproducible voice messages and the quality of their sound. To do this, when accumulating spectroband oscillatory components allocated by vocoder channel-forming filters, by registering these components in the memory of a memory device, the oscillatory components are divided into impulse responses of the module logarithm and sign. When synthesizing speech signals, pulse response counts are enhanced by summing up with control signals, and before playing, they expand and assign previously separated characters. 2c. and 1 z. p. f-ly, 1 ill.

(L

00

N (cd

ten

15

20

11408450

The invention relates to speech informatics, in particular to digital speech autoinformers, and can be used in computing and communication technology to reproduce quantized vocoder components.

The purpose of the invention is to improve the dynamic properties of reproducible voice messages and the quality of their sound.

The goal is achieved by the accumulation of spectroband oscillatory components, selected by channel-forming filters of the vocoder, by registering these components in the memory memory device, the oscillatory components are divided into impulse responses of the module logarithm and sign, and when synthesizing speech signals, the impulse response counts are enhanced by summing with control signals, and before playing the expander, KiT and assign previously separated characters.

The drawing shows a structural electrical circuit of the proposed device.

The device for synthesizing speech i signals contains a memory 1 with an address generator 2, the input of which is connected to a clock generator 3, a separator 4 signal symbols connected by an input to the output of a memory device, and the outputs are connected to inputs of a digital adder 5 signals. The output of the digital adder is connected to the address inputs of the persistent storage device a, on which the expander 6 is made. The output of the expander is connected through accumulation of the adder 7 to the input of the digital-analog converter 8, the output of which is the output device for speech synthesis, and the sign input of the accumulating adder connected via the corresponding output of the discriminator of the signal symbols to the output of the memory device. Address generator 2 is implemented as a serial connection of the shift register 9, the address generator of this address generator. The logical block of the address generator is also connected by inputs to the corresponding inputs of control signals and outputs to the installation inputs of the shift register 9.

The device implements the proposed method as follows.

At the beginning of the output of voice messages, control signals arrive at the inputs for setting the states of counters, registers and triggers, the pulses of the clock generator cause further changes to these states and addresses transmitted to the address inputs of the memory 1. The characters read from the memory are decrypted into the separator 4 signal symbols, what is related to the current levels of bandwidths stored in the memory, is summed with the control signals that determine the levels of oscillations in the bands; a digital adder 5. The output of a digital adder serves as the address of a persistent storage device that performs the functions of an expander with respect to the instantaneous values of bandwidth; The expander output 6 is then combined with the output of the signal symbol separator, which transmits information about the current oscillation sign. All this enters accumulating adder 7. The operation of the clock generator 1 through the address generator 2 leads to a signal sweep along the bands and periods of the pitch period, if control signals indicate that the sound was localized, otherwise the frequency of the time base is replaced by the quasi-random nature of the changes in the addresses of the information read by the memory device. and, at the output of the analog converter, reproduction of the sequence of samples imitating the noisy sounds of speech. With a periodic scan of addresses, a periodic sequence of samples is formed, which is determined both by the contents of the storage device and by all the control signals received 30

35

40

45

unit 10 and unit 11 counters to the control input of the device

cove The outputs of the counters are digitally synthesized output addresses of the address generator, and the logged compression of the shift register is the input to the vocoder's signals.

ten

15

0

0

this address generator. The logical block of the address generator is also connected by inputs to the corresponding inputs of control signals and outputs to the installation inputs of the shift register 9.

The device implements the proposed method as follows.

At the beginning of the output of voice messages, the control signals arrive at the inputs for setting the states of counters, registers, and triggers, the pulses of the clock generator cause further changes to these states and addresses transmitted to the address inputs of the memory 1. The characters read from the memory are decrypted in the separator The 4 signal symbols, which refer to the current levels of bandwidths stored in the memory, are added to the control signals that determine the levels of oscillations in the bands; n is a digital adder 5. The digital adder output location is constant storage device performs the function of the expander guide relative to the instantaneous values of m oscillations lane; The expander output 6 is then combined with the output of the signal symbol separator, which transmits information about the current oscillation sign. All this enters accumulating adder 7. The operation of the clock generator 1 through the address generator 2 leads to a signal sweep along the bands and periods of the pitch period, if control signals indicate that the sound was localized, otherwise the frequency of the time base is replaced by the quasi-random nature of the changes in the addresses of the information read by the memory device. and, at the output of the analog converter, reproduction of the sequence of samples imitating the noisy sounds of speech. With a periodic scan of addresses, a periodic sequence of samples is formed, which is determined both by the contents of the storage device and by all the control signals received by

five

0

five

Claims (3)

Claim 1. A method for synthesizing speech signals accumulated by a digital method with 15 minimum redundancy by generating control signals for sampling the sampled logarithmic spectral components of speech signals and then converting them 20 to analog ones, which involves improving the dynamic properties of reproducible speech, messages, when sampling discretized logarithmic spectral components, the responses of the storage device are divided into pulse responses of the logarithm of the module and the sign the first one, moreover, the response samples are amplified by summing the pulse responses of the logarithm of the module with control signals, and then expanding and assigning previously separated characters. 2. A device for synthesizing speech signals accumulated by a digital method with minimal redundancy, comprising a memory device with an address generator, to the input of which a clock generator is connected, as well as a signal symbol separator connected by an input to the output of the memory device, and a digital-to-analog converter, the output of which is the output of the device characterized in that, in order to improve the quality of the displayed audio messages, an accumulating adder and an expander are introduced into it, moreover, the sign output The signal symbol separator is connected to the corresponding input of the accumulating adder, the remaining inputs of which are connected to the expander outputs, and the outputs are connected to the inputs of the digital-to-analog converter. 3. The device pop.2, characterized in that the expander is made on a permanent storage device, and the address generator is made by the serial connection of the shift register, logical block and block counters.