SU142698A1 - A device for separating articulation signal parameters and a speech signal residual signal - Google Patents

Description

Devices for separating articulation signal parameters and a speech signal residual signal are known.

In the described device, in contrast to the known ones, the articulation signal-parameters and the signal-residue of speech from speech messages are extracted and the speech is restored at the receiving end, but not. imitation of voice messages. Linear contour with variable parameters is excited not by a harmonic noise generator, but by a voice. The signals at the output of statistical sensors are always close to the reference values, and the control signals are variable time functions determined by the parameters of the speech signal.

The removal of redundant information during voice transmission occurs because at the transmitting end of the device a linear circuit with variable parameters is used to isolate the signal-residual connected through a matching unit, a sensor of current statistical parameters and gain controllers included in the feedback circuit to control the operation. linear circuit, and at the receiving end for synthesizing the speech signal a linear circuit is applied, inverse to the characteristics of the transmitting circuit and the controlled signal parameter and speech, obtained at the transmitting end.

The block diagram of the transmitting part of the described device is shown in FIG. one; the block diagram of the receiving part is shown in FIG. 2

The transmitting part of the device is a system with a complex feedback. It uses linear i loop /, with variable parameters, stimulated by voice message z (f), which is fed to input 2 to separate the articulation signal parameters and the speech signal balance.

No. 142698- - 2 -

statistical parameters of the analyzed random function x (t) through the device 4 matching the circuit with the sensor. In the feedback loops, the peaking devices 5 are turned on. When the gain is sufficiently high, a balance is established in the feedback loop. The series of statistical parameters of the signals Xi (t) and Xi (t), which are auto-controlled, is kept constant. At the same time, information is filtered out of the speech signal z (t) and extracted as a signal parameter t / i (/) .... yn (i) at the outputs 6. The signal Xi (t) obtained at the output 7 of the linear circuit 1 Is a residual signal and is not transmitted to the line.

At the receiving end, the signal-residue Xi {t} is simulated by a local source and fed to input 8. To restore the original speech signal z {t), a linear circuit with variable parameters 9 is used, with a characteristic opposite to the characteristic of the analyzing circuit. Linear circuit 9 is controlled by speech signal parameters yi (t) Yn ((), arriving at inputs 10, and excited

signal balance. The output is the original speech signal z (t). The described device can be used in communication technology.

Subject invention

A device for extracting signal signaling parameters and a speech signal residual signal, characterized in that, in order to eliminate redundant information during voice transmission, a linear circuit with variable parameters is used at the transmitting end to separate the signal residual connected through a matching unit sensor current statistical parameters and gain controllers included in the feedback circuit to control the operation of the linear circuit, and at the receiving end, to synthesize the speech signal, a line is applied A circuit that is inverse to the characteristics of the transmitting circuit and controlled by the signal-parameters of speech obtained at the transmitting end.

7 lx, ft)

x, (tj

 6 V

r

y, y / "

ig. 2