JPH082047B2 - Confidential device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a secret talk device, and more particularly to a secret talk device that transforms a voice signal at a transmission side and restores it at a reception side.

[Conventional technology]

What is demanded of the confidential device is high quality and high confidentiality of sound quality under the limitation of the transmission capacity of a given transmission line such as a public communication telephone line, and there is a trade-off relationship between them.

The transformation process and the restoration process of the audio signal for the confidential talk are linear arithmetic processing, and when highly confidentiality is required and complicated processing is required, for example, block-wise processing such as FFT is involved.

[Problems to be Solved by the Invention]

In a conventional confidential communication device involving block-like complicated processing, for example, when a voice signal having a waveform shown in FIG. 2 (a) is transformed, transmitted, and restored, due to restrictions on the amount of calculation and non-linearity of the transmission path. Since there is a limit to the reproducibility of the waveform, Fig. 2 (b)
As shown in, the restored audio signal has a waveform discontinuity at a block boundary, resulting in a poor sound quality.

An object of the present invention is to provide a secret-speaking device having good sound quality without block discontinuity at a block boundary in a restored voice signal while performing block-like processing to obtain high secrecy.

[Means for solving the problem]

The secret-speaking device of the present invention is a secret-speaking device that scrambles a voice waveform by block-like processing and sends it as a secretized waveform from a transmission side to a reception side, and first means for calculating a linear prediction coefficient of a voice signal, Second means for inverse filtering the audio signal using the calculation result input from the first means, and third means for removing a predetermined frequency section of the signal input from the second means, A fourth means for expressing the calculation result input from the first means within the predetermined frequency section;
And means for synthesizing the signals input from the third means.

〔Example〕

 Next, the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a first embodiment of the present invention.

The embodiment shown in FIG. 1 is an A / D converter, a Hamming window 2, an autocorrelation calculator 3, an LPC analyzer 4, an LPC inverse filter 5, a band elimination filter 6, a conversion circuit 7, a synthesis circuit 8, a scramble circuit. 9, D / A
Converter 10, A / D converter 11, descramble circuit 12, separation circuit 13, inverse conversion circuit 14, LPC synthesis filter 15 and D / A converter 16
It is configured with.

The syllable signal Si to be transmitted is a signal whose band is limited to 4 KHz or less, and the sampling frequency is 8 KH by the A / D converter 1.
It is sampled at z and quantized to the required number of bits.

The Hamming window 2 has a window length of 30 ms for the output signal of the A / D converter 1.
The window processing for multiplying the Hamming window function of is performed every 20 ms period (repetition frequency is 50 Hz). The autocorrelation calculator 3 calculates the autocorrelation coefficient sequence of the signal waveform blocked by the Hamming window 2. The LPC analyzer 5 uses this autocorrelation coefficient sequence to perform linear prediction of each block of the signal waveform.
The α parameter which is a Coding (LPC) coefficient is calculated.

LPC with the α parameter output from the LPC analyzer 4 as a coefficient
The inverse filter 5 inversely filters the output signal of the A / D converter 1 and outputs the residual signal of the LPC analysis by the LPC analyzer 4.

The band elimination filter 6 removes the components in the frequency section 1500 to 2125 Hz from the residual signal input from the LPC inverse filter 5 and outputs the other components. It should be noted that 1500 to 2125 Hz is selected as a frequency section in which the removal of components has little influence on the sound quality.

The conversion circuit 7 outputs a signal waveform that expresses the α parameter input from the LPC analyzer 4 in the frequency section 1500 to 2125 Hz.

The synthesis circuit 8 synthesizes the signals input from the band elimination filter 6 and the conversion circuit 7.

The combined signal output from the combining circuit 8 is scrambled on the frequency axis by, for example, FFT scrambling in the scramble circuit 9, is analogized by the D / A converter 10, and is then sent to a transmission line (not shown). It

The signal transmitted to the receiving side through the transmission line is the A / D converter 1
It is digitized by 1 and descrambled by the descramble circuit 12. The separation circuit 13 separates the signal input from the descrambling circuit 12 into a component in the frequency section 1500 to 2125 Hz, which is a waveform representation of the α parameter, and a residual signal excluding the components in this frequency section. Inversion circuit 14
Performs the inverse conversion of the conversion by the conversion circuit 7, and the separation circuit 13
The signal waveform consisting of the components in the frequency range 1500 to 2125 Hz input from is inversely converted to the α parameter. When this α parameter is used as a coefficient, the LPC synthesis filter 15 inputs the residual signal from the separation circuit 13 and synthesizes a digital voice signal, and this digital voice signal is analogized by the D / A converter 16 and the voice signal Output as S ₀ .

If the waveform of the residual signal output from the LPC inverse filter 5 on the transmitting side is as shown in FIG. 3 (a), the residual signal restored on the receiving side and output from the separation circuit 13 is
As shown in FIG. 3B, the waveform has discontinuity at the block boundary. However, this discontinuity is caused by the LPC synthesis filter 15
Since the output voice signal S ₀ is smoothed by the filtering, the waveform of the output voice signal S ₀ is smooth even at the boundaries of the blocks as shown in FIG.

 The embodiment shown in FIG. 1 has been described above.

In the embodiment shown in FIG. 1, the speech signal Si is separated into an LPC coefficient, which is spectral envelope information, and a residual signal, which is spectral fine structure information, and then all processing is performed at the waveform level. The second embodiment described below is an embodiment in which most of the processing is performed at the spectrum level.

FIG. 4 is a block diagram showing a second embodiment of the present invention. However, only the transmitting side is shown and the receiving side is omitted.

The embodiment shown in FIG. 4 has an A / D converter 1, a Hamming window 2, an autocorrelation calculator 3, an LPC analyzer 4, an LPC inverse filter 5, a rectangular window 17, D.
FT circuit 18, band remover 19, LPC analyzer 20, interpolator 21, frequency converter 22, synthesis circuit 23, frequency replacement circuit 24, IDFT circuit 25, D /
It is provided with an A converter 26 and a receiving side (not shown).

The LPC analyzer 4 outputs the α parameter and the LPC inverse filter 5 outputs the residual signal, as in the embodiment shown in FIG.

The rectangular window 17 blocks the residual signal input from the LPC inverse filter 5 every 32 ms (repetition frequency is 31.25 Hz) by rectangular window processing.

The DFT circuit 18 transforms the signal waveform blocked by the rectangular window 17 into a frequency spectrum by performing a discrete Fourier transform at (8000 / 31.25 =) 256 points. The band remover 19 is a frequency section 1500 to 2125 Hz of the frequency spectrum input from the DFT circuit 18.
The spectral density at is output as 0.

The LSP analyzer 20 uses the LSP (line sp
ectrum pair) Convert to (column of). The LSP coefficient output from the LSP analyzer 20 is interpolated at an interval of 31.25 Hz by the interpolator 21, and the interpolated LSP coefficient of 0 to 4 KHz is the frequency converter 2
The frequency is converted into the frequency range 1500 to 2125 Hz by 2.

The combining circuit 23 combines the signals input from the band remover 19 and the frequency converter 22.

The combined signal output from the combining circuit 23 is the frequency replacement circuit 24.
In the IDFT circuit 25, the inverse Fourier transform is applied to the frequency replacement signal, and the signal in the frequency domain is converted into the signal in the time domain, in other words, the signal waveform is converted into an analog signal with the D / A converter and the transmission line is converted. Sent to.

On the receiving side, the band remover 19
The residual signal and the LSP coefficient (interpolated by the interpolator 21) are restored by the LSP synthesis filter using the restored LSP coefficient, and the restored residual error A voice signal is synthesized from the signal. The waveform discontinuity at the block boundary of the restored residual signal is LS
It is the same as in the embodiment shown in FIG. 1 that the sound quality of the synthesized voice signal is improved as a result of being smoothed by the P synthesis filter.

〔The invention's effect〕

As described above, the present invention provides the LPC by the first means.
A coefficient, the second means obtains a residual signal, the third means removes a specific frequency section of the residual signal, and the fourth means expresses the LPC coefficient in a specific frequency section. By combining the output signals of the means and the fourth means by the fifth means, the residual signal and the LPC coefficient are restored from the synthesized signal output by the fifth means, and the synthesized filter using the restored LPC coefficient is used. By synthesizing the voice signal from the residual signal, it is possible to eliminate the discontinuity of the waveform at the boundary of the signal processing block, and as a result, it is possible to improve the sound quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing the first embodiment of the present invention, and FIG.
FIG. 4 is a waveform diagram for explaining the problems in the conventional confidential communication device, FIG. 3 is a waveform diagram for explaining the operation of the embodiment shown in FIG. 1, and FIG. 4 is a second embodiment of the present invention. It is a block diagram showing. 2 ... Hamming window, 3 ... Autocorrelation calculator, 4 ... LPC
Analyzer, 5 LPC inverse filter, 6 band elimination filter, 7 conversion circuit, 8 synthesis circuit, 13 separation circuit, 14 inverse conversion circuit, 15 LPC synthesis filter, 17 …
… Rectangular window, 18 …… DFT circuit, 19 …… Band eliminator, 20 …… L
SP analyzer, 21 ... Interpolator, 22 ... Frequency converter, 23 ...
Synthesis circuit.

Claims (1)

[Claims] Claims: What is claimed is: 1. A secret-talking device for scrambling a speech waveform by block-like processing and sending it as a confidential waveform from a transmitting side to a receiving side, comprising first means for calculating a linear prediction coefficient of a speech signal, and A second means for inversely filtering the audio signal using the calculation result input from the first means; a third means for removing a predetermined frequency section of the signal input from the second means; A fourth means for expressing the calculation result input from the first means within the predetermined frequency section and a fifth means for combining the signals input from the fourth means and the third means. A secret-talking device characterized by being equipped.