SU372720A1 - SYNTHESIZER OF NONLINEAR HARMONIC VOCODER - Google Patents

Description

il

The invention relates to synthetic telephony and can be used in devices for transmitting voice messages in a narrow frequency band using the harmonic coding method.

Known synthesizers are a cascade of linear harmonic synthesis. In linear harmonic vocoders, a level-independent error is an absolute error, as a result of which spectral components are synthesized with low levels with unacceptable errors, and with high levels with excessive accuracy.

In order to improve the quality of the synthesized speech, in the proposed synthesizer of a non-linear harmonic vocoder, equal angles are equal, the equal inputs of linear harmonic synthesizers are connected in parallel. The output of the first linear harmonic synthesizer.

The drawing shows a block diagram of a device for non-linear harmonic coding of speech with the proposed synthesizer.

The microphone / is connected to the comb of bandpass filters 2 of the spectrum analyzer and to the main tone extractor 3. The outputs of the no-band filters through detectors 4 are connected to devices for extracting the root 5, which, in turn, are connected to the harmonic matrix 6. The outputs of the matrix are connected to control inputs 7 of the first linear harmonic synthesizer containing the multiplier 8 and the delay line 9, and in parallel with the like inputs

second linear harmonic synthesizer containing multipliers 10 and delay line //. The output of the first synthesizer is connected to the drive input 12 of the second synthesizer, the output 13 of which is the output of the device.

A speech signal with an instantaneous spectrum of 5 (co) from a microphone / is fed to a comb of t bandpass filters 2. The output voltages of filters 2 are detected by detectors 4, resulting in t voltage values that are proportional to the envelope samples of the central frequencies filters 2. The outputs of devices 5 receive a signal representing the square root of the spectrum, close to the logarithm of the spectrum according to the nature of the nonlinear transformation. The matrix 6 determines the coefficients of the decomposition of the incoming voltages.

Thus, the pattern of the outputs of the matrix is proportional to the coefficients of the square root decomposition of the spectrum of the original speech in the Fourier series.

The resulting voltages are supplied as modulating signals to multipliers.

8 and 10 synthesizers, in which they are multiplied by the excitation signal taken from the output of the excitation generator 14. The generator 14 is controlled by the pitch signal from the selector 3.

The signals from the outputs of the multipliers arrive at the taps of the delay lines, closed at one end and consistently loaded at the other. The signal from the output of the delay line 9 is fed to the inputs of the multipliers 10 of the second synthesizer. The signal spectrum at the output of the delay line // is equal to the spectrum of the input speech signal, which can easily be proved mathematically.

The invention makes it possible to redistribute the accuracy of the amplitude approximation of the spectrum by increasing the accuracy by

low levels by reducing it at high levels and thereby improving the quality of the synthesis.

Subject invention

A nonlinear harmonic vocoder synthesizer containing two linear harmonic synthesizers, each of which has control inputs and an excitation input, characterized in that, in order to increase the quality of the synthesized speech, the same control inputs of the linear harmonic synthesizers are connected in parallel, and the output of the first linear harmonic synthesizer connected to the excitation input of the second linear harmonic synthesizer.