Method for creating a speech-like masking signal

FIELD: technology for creating artificial interference for masking acoustic and electromagnetic information leakage channels.

SUBSTANCE: speech-like signal is composed of original signal being masked by modulation with noise signal of moments when signal being masked crosses zero. Modulation of zero crossing moments of speech signal may be performed by phase modulation of speech signal by noise signal. Modulation of zero crossing moments of speech signal may also be performed in accordance to a different algorithm - speech signal is clipped, resulting clipped signal is integrated, resulting saw-shaped signal is added to noise signal, after that resulting digital signal is clipped. For masking speech signal in digital form, resulting masking signal is added to the same signal converted to digital form. Modulating signal is a noise signal. Control listening indicates that resulting signal is unintelligible, although it preserves the speech-like form.

EFFECT: increased masking effect of speech-like noise signal.

3 cl, 2 dwg

The technical field.

The invention relates to the field of artificial noise to mask acoustic and electromagnetic leakage channels of voice information.

The level of technology.

It is known that the value of economic, political, scientific and technical information exceed the value of material objects. It is therefore extremely urgent protection from unauthorized leaks of information from the state, defense, scientific-technical and commercial content.

In addition to organizational measures, prescribing the staff rules, to protect the content information apply sophisticated technical means. In particular, to reduce the leakage of information on acoustic and radio channels, a telephone network, a circuit used equipment, etc. applied special schemes masking information signals, reducing the possibility of unauthorized listening (for example, Ratkevich Century, Ratkevich P. "Technique measurements at the radio. M: Communications, 1969, str-381).

In operation Horev A.A. ("Methods and search tools, electronic devices intercept information". M: the defense Ministry, p.24) presents examples of network bookmarks that are installed in electrical outlets, extension cords, equipment, mains operated displaced the nogo current or directly in the power line. Bookmarks are used for retrieval of verbal information. Equipment gain of verbal information should be protected by the channel leak power. Without protective measures amplifying equipment makes a reaction in the network in accordance with the amplified signal, which allows their separation from the network, i.e. the possibility of information leakage. Use power strips and equivalents network in this frequency range is almost not impair the electrical signals. Network filters on the audio frequency bulky and provide in the range of 100-10000 Hz the attenuation is of the order of 5-10 dB, which is clearly not enough.

Other information security technical means is the creation of artificial interference in the electromagnetic channels of information leakage generators masking signal power exceeds the rated power of the electromagnetic signals of information leakage. This significantly reduces the possibility of detecting and receiving information. Modern modification of this method of masking is shown, for example, in the application Vselennaya No. 2005103317 February 10, 2005 (the Way the masking electromagnetic leakage channels of speech signals, sound equipment). The method provides for the generation, amplification and radiation of the masking noise signal, and is characterized by the fact that the strengthening of omologo signal produced in mode b or AB, apply a pulse autocompensation the voltage of the power amplifier of the noise signal, when filtered stabilized voltage by the high-pass filter.

The masking noise signal in the transmitting antenna (acoustic, electric, magnetic). The power output of the noise signal must substantially exceed the signal power leakage of the masked signals.

The method effectively prevents or substantially reduces electromagnetic information leakage and air, and industrial AC.

Currently, however, there are many effective ways (e.g., correlation), allowing you to select the speech signal even in the face of strong noise. The first step for the organization of eavesdropping is to identify the presence of a signal in the analyzed mixture signal and masking noise. If you know that this mixture has a signal - sooner or later, the signal will be selected.

Development of a method of masking is the selection of the masking signal. To ensure secrecy (exclude the possibility of eavesdropping) is necessary: either to radiate an extremely powerful speech and music signals, suppressing the signals of acoustic and electromagnetic leakage of masked speech signals (signals Kahn is the fishing information leakage are incomparably less power) or as to suppress the signs of speech of the possible leakage of the masked speech signals, which requires extremely large capacity masking noise and, nevertheless, does not exclude the possibility of the detection of the speech signal.

Another disadvantage of using the noise signal as a mask is that its peak-factor criterion 3σ 9.5 dB (voltage), while the peak factor of speech is 14 dB. To suppress the speech signal it is necessary to increase the level of the noise signal so as to cover the level of the speech signal. The increase in the level of the noise signal destroys to a large extent the zero-crossing of the speech signal, i.e. the signal in the noise virtually impossible to detect. Measure the destruction of the speech signal is the destruction of the zero-crossing to the extent that the law becomes almost normal, because you know that deep amplitude limiting speech signal its intelligibility remains high.

There is a method of masking by radiation in the ether false distracting speech signals. To increase the reliability of masking false distracting speech signal can be performed Maarsbergen. Listening party in the end will find it among the many other signals and nacket actively analizirovat the ü. It will distract her from seeking real masked source, thus time will ensure its secrecy (Weggemann, Innoko, Ivina. "Digital steganography. Aspects of protection". M: SALT press, 2002).

The most known way to create speech-like noise signal - mosaic frequency and temporal permutation of the components of the speech signal. He used, for example, in the scrambler "GROTTO", described in "Directory MASK" center of information security, ed. 2001

It provides for the separation of the current speech spectrum or time interval of the speech signal in the frequency or time slices, and mutual rearranging them in a specified manner. This method is most similar to the claimed and therefore chosen for the prototype.

The method has been used successfully for many years, allows you to make voice signal unintelligible in the chains of transmission and the possible channels of information leakage, but it can be completely restored at the receiving end.

However, when applying this method remains the principal possibility to restore the original signal to the listening party, for example, by the disclosure of the algorithm permutations of the elements of the speech signal by means of computer processing of suspicious recorded fragment or theft of the processing algorithm. This may partially correct the th periodic change in the algorithm, but it does not guarantee the secrecy of speech information since the time of disclosure of the algorithm listening party unknown masked side.

In addition, the speech is converted into digital form, poorly masked noise signal as the signal in digital form is complex, and the noise signal is simple.

Furthermore, the method-prototype roads, because it requires the creation of tunable mask encoding and decoding devices and the rule of secrecy coding on all objects.

The objective of the invention.

The objective of the invention is to enhance the masking effect of speech-like noise signal and reduce the cost of its implementation. It is necessary to generate the masking signal, which masks the analog speech signal and the speech signal in digital form.

The inventive solution of the task.

The original background.

Theorists of speech intelligibility showed that the intelligibility of distorted speech depends mainly on maintain the intersection points of the speech signal zero (mean point of the waveform of an electrical signal, the mutual position on the time axis of the times at which the waveform crosses the time axis. In other words, the point of zero crossing is the moments of time at which the instantaneous value of the signal is avno zero). Therefore, clipped speech (i.e. speech, maximally enhanced and limited in amplitude remained unchanged when only the point of zero crossing) is quite picky and even recognizable.

It is also known that the point of zero crossing are some of the main characteristics of the formant speech. So the above idea can be expressed differently, the speech intelligibility is determined by the preservation of its formant structure (Rquotad. Voice control robot". M.: Radio and communication, 1989).

From the popularity of these facts the applicant has suggested an inverse relationship - you can distort it to a full degree by destroying its formant structure. This can be accomplished by modulation of the noise signal points of zero crossing of the speech signal.

It is known that the phase jitter of the speech signal randomly in communications equipment, magnetic recording cause frequency modulation as harmonic and pulse signal (Butler I.M., Driatzkaya I.N. "Digital transmission of audio broadcasting. M.: Radio and communication, 1989), reducing the quality vosproizvedennogo speech signal, i.e. its intelligibility. In the work Vimana GV ("voice Transmission over telecommunications networks". - M.: Radio and communication, 1985, s) are called the formants enhanced frequency domain spectrum of this sound, to distinguish it when SL is fur perception from other sounds. The speech sounds differ in their spectral composition from each other as the number of formants and their location in the frequency spectrum. Under the range of formant refers to the dependence neveroyatnye for a long time spectral formant level At the frequency f.

In the work Potapova R.K. (Voice control robot". - M.: Radio and communication, 1989, p.9) defined the concept of frequency formants as the maximum spectral density of the speech signal or as the maximum average frequency spectrum in the region of the respective formants determined by the number of transitions per unit time in the respective frequency range. One of the ways to select the formant is to determine the maximum number of intersections of the speech signal through zero in the corresponding frequency domain by measuring the density of transitions in the speech signal. The destruction of the density of transitions of the speech signal in the corresponding frequency domain reduces the intelligibility of speech.

In the work Baburkina V.N., Hansel GS and Pavlov N.N. ("Elektroakustische and broadcasting. Acoustic issues of broadcasting". - M.: Communication, 1967, s) shows that the parasitic frequency modulation in the form of detonation distorts the sound signals. Parasitic frequency modulation distorts the spectral distribution of the sound signal (Kantor L.Y. "Methods of increasing omehow is tasciotti receive FM signals. M: Communication, 1967). In the work of many CENTURIES ("Measurement equipment audio recording movies". M: Art, 1962, p.56-63) presents the threshold levels of hearing sensitivity to parasitic frequency modulation. They are from 0.003% to 0.03%, depending on the frequency modulated oscillations and the modulation frequency. The increase of the parasitic frequency modulation reduces the intelligibility of speech.

The problem is solved by the fact that in the known method of forming speech-like masking signal generated from the original speech signal, significant changes were made, namely:

- speech-like signal created from the original speech signal by phase modulation noise signal that leads to the destruction of its formant structure by changing the points of zero crossing of the original speech signal.

Phase modulation of the original speech signal noise signal randomly changes the moments of the zero crossing of the original speech signal.

Chaotic change (modulation) of the moments of the zero crossing of the source speech signal noise can also be slightly different before the phase modulation of the original speech signal Klippert, and then clipped speech signal is subjected to phase modulation. In this case, the phase modulation can be performed, for example, as follows. Clipped signal SN is Chala integrate, converting into sawtooth, then the sawtooth signal is summed with the noise signal, then the received total signal increase and Klippert.

For masking speech signal in digital form of the received speech-like masking signal is summed with the signal converted into digital form.

Disclosure of the invention.

The method explain 1 and 2, where:

figure 1 - example of a structural diagram of the method;

figure 2 - chart of the stress diagram figure 1.

In scheme 1, the following notation: 1 - source speech signal, 2 is the noise source, 3, 6, 7 filters low pass, 4, 8 Comparators, 5 - integrator, 9 - a / d Converter, 10 - adder.

In figure 2 the following notation:

and the form of the original speech signal at the output of source 1;

b - form of the output signal of the lowpass filter 3;

in the form of the output signal of the comparator 4;

g - form of the signal at the output of the integrator 5;

Dr. waveform at the output of the lowpass filter 6;

e - the shape of the signal at the output of the comparator 8.

The simplest implementation of the method (figure 1) consists of a comparator 4, to the first input through a low-pass filter 3 and a decoupling capacitor (to avoid constant component) connected to a source 1 of the speech signal, the second no load) the e constant voltage, the output of the comparator 4 through the integrator 5 is connected to the first (+) input of comparator 8. The source of the noise signal 2 through the filter 6 of the lower frequencies is connected to the reference (-) input of comparator 8. The output of the comparator 8 is connected through the filter 7 of the lower frequencies to the input of the analog-to-digital Converter 9. The outputs of the comparator 8 and the analog-to-digital Converter 9 through the adder 10 is connected to the output device.

The method of figure 1 is implemented as follows. The signal from the source speech signal (figa) through the filter 3 of the lower frequencies is fed to the first (+) input of comparator 4. When applying the speech signal through a lowpass filter with a cutoff frequency of 2000 Hz signal to the input of the comparator 4, would be figb. On the reference input of the comparator 4 is zero potential. At the output of the comparator 4 is formed clipped speech signal, i.e. the speech signal is amplified and clippered into a sequence of rectangular pulses (pigv). Then clipped signal is integrated by the integrator 5, converting the rectangular pulses in a sawtooth (Figg). Received sawtooth pulses are received on the first (+) input of comparator 8. At the same time on the reference (-) input of the comparator 8 through the filter 6 of the lower frequency signal from the noise source 2 (high). This noise signal is continuously shifts the threshold comp is atora 8, in the result at the output is formed clipped signal, the fronts of which is delayed (shifted in time) relative to the points of zero crossing of the original analog speech signal, and the delay of each subsequent front chaotically changing in accordance with the signal noise (figd). Analog-to-digital conversion of the received mask signal produced by the block 9, forms a digital signal, masking the original speech signal in digital form. The lowpass filter is required to limit the spectrum of the signal before conversion.

In the adder 10 is formed joint speech-like noise masking at the same time the speech signal in analog and in digital form.

The intelligibility of speech violated violated because the formants of the original signal as the moments of the zero crossing of the source speech signal modulated by a noise signal.

It is also possible implementation of the method through direct phase modulation of the original speech signal.

Industrial applicability.

The claimed method is repeatedly implemented by the applicant, in particular, according to the scheme 1.

While there were no technical difficulties. Applied elements (Comparators, integrators, filters, a lowpass analog-to-digital Converter, an adder known for a long time, DOS is publicly accessible and does not require any modifications. The shaper is inexpensive and compact.

Control listening formed signal affirmed its full promiscuity while maintaining receptonist.

Thus, in our opinion, the claimed method meets the criteria of the invention - it is new, not obvious and industrially applicable.

1. The method of formation of speech-like masking signal generated from the original speech signal, wherein the speech-like signal created from the original speech signal by phase modulation noise signal, which leads to the destruction of the formant structure of the original speech signal.

2. The method according to claim 1, characterized in that before the phase modulation of the original speech signal Klippert, and to perform a phase modulation clipped the original speech signal of the first integrating, transforming it into a sawtooth, then the sawtooth signal is summed with the noise signal, then the received total signal increase and Klippert.

3. The method according to claim 1 or 2, characterized in that the speech-like masking signal is summed with the same signal, converted into digital form.