EP0044845B1 - Process and device for protecting the transmission of vocal signals - Google Patents

Abstract

Process and device for protecting the transmission of vocal signals along a limited band-width analog transmission channel (for instance a vocal CCITT channel), protecting against illicite listening. By exploiting the speech information perceived by the human ear, the vocal signal to be transmitted is digitalized and undergoes a corresponding information reduction (A(f)) based on Fourier's transformation. The necessary information protection can thereby be obtained in the subsequent coding, with respect to the transmission distortions and the modulation is optimally adapted to the transmission behaviour. The digital form existing before this coding can be easily digitalized so that the transmitted analog signal is totally incomprehensible and can not be reconstituted even through the modem methods of analysis, while for the authorized listeners a good intelligibility subsists. By this process, through the modern means of the digital signal processing technique, an optimal solution to the safe analog transmission of speech can be achieved.

Description

The present invention relates to a method for the secure transmission of voice signals according to the preamble of claim 1 and a circuit arrangement for carrying out the method according to the preamble of claim 6.

As is well known, methods of protecting against unauthorized eavesdropping are increasingly being used in the transmission of voice messages. The corresponding voice transmission takes place for this purpose either by: coding or parameterization with subsequent encryption or by concealment. To date, speech concealment devices can be divided into six development generations, the last no longer starting from a transmission of the speech signal itself, but from an orthogonal transformation of the speech signal (cf., inter alia, "I EEE Transactions on Audio and Electroacoustics", vol. AU-21 , No. 3, June 1973, pp. 165-174, and P. Meier, _" Secure Speech Communication over a CCITT-Speech Channel", published in _" NATO ASI Proceedings on Speech Understanding and Synthesis, _" Bonas France, July 1979) .

A method for concealing voice signals according to the preamble of claim 1 is known (DE-B1 No. 2523828). In an arrangement for carrying out this method, a speech signal limited in bandwidth to 3 kHz is pulse-code modulated and divided into samples controlled by a time scrambler. Image samples are obtained from the samples using an orthogonal matric and their values are successively transmitted in sections from a transmitter to a receiver.

In theory, the transmission path has been assumed to be ideal in this method; In practice, only a very limited intelligibility can be achieved, since the system limits of a real transmission path (e.g. CCITT voice channel) are not sufficiently taken into account.

Furthermore, a vocoder system for concealed transmission of voice signals is known from US-A No. 4064363, in which the input signal is divided into a baseband and an upper band with regard to its frequency spectrum on the transmitter side. Following a discrete Fourier transformation of the signal, all of the Fourier coefficients containing the amplitude and phase information and only quantity-related, derived quantities from the baseband are further processed together in a common time-multiplex frame and transmitted to the receiver side via a modem (not described further). While the conditions of the transmission path are already taken into account here, the security against unauthorized unveiling is still low.

It is therefore the object of the invention to provide a method and a circuit arrangement for secure analog and narrowband voice transmission which, while maintaining a high voice quality at the authorized recipient, at the same time enable substantially increased security against unauthorized eavesdropping.

The object is achieved in a method and a circuit arrangement of the type mentioned by the features of the characterizing part of claim 1 and claim 6, respectively.

The essence of the invention is to take advantage of a known information reduction, which is achieved by omitting certain phase information in a frequency range above about 1 kHz which is insignificant for human hearing, in order to additionally use an image sample obtained by the orthogonal transformation despite a limited bandwidth Secure encryption transmitted and reduce the residual intelligibility by weighting the spectrum. On the one hand, this results in greater protection against unauthorized eavesdropping and, on the other hand, a reduction in the interference sensitivity of the signals in the transmission path.

The method according to claim 1 can be implemented relatively easily with the means of modern electronics and has the advantages of high operational and interference immunity and good transmission quality.

The advantage of a method expanded according to claim 3 lies in the additional security against eavesdropping achieved.

The method and a preferred circuit arrangement are explained in more detail below with the aid of drawings; The components of the system which are not used to secure unauthorized eavesdropping are not shown in detail for the sake of simplicity.

• Fig. 1 ein Blockschaltbild eines Senders,
• Fig. 2 ein Blockschaltbild eines Empfängers in einem System zur sicheren Sprachübertragung,
• Fig. 3 ein abgetastetes, pulskodemoduliertes Sprachsignal von 20 ms Dauer,
• Fig. 4 das komplexe diskrete Fourierspektrum des Signals gemäss Fig. 3,
• Fig. 5 das Fourierspektrum gemäss Fig. 4 nach seiner Gewichtung,
• Fig. 6 eine Darstellung resultierender Sprachparameter nach einer Informationsreduktion und dreifachem Duplizieren,
• Fig. 7 die Sprachparameter gemäss Fig. 6 nach ihrer Permutation,
• Fig. 8 die Sprachparameter gemäss Fig. 7 nach ihrer Kodierung,
• Fig. 9 das übertragene modulierte Sprachparametersignal (Sendersignal f' (t)),
• Fig. 10 die Sprachsignalparameter nach ihrer Demodulation im Empfänger (vergleichbar mit Fig. 8),
• Fig. 11 die Sprachsignalparameter nach ihrer Dekodierung (vergleichbar mit Fig. 7),
• Fig. 12 die Sprachsignalparameter nach ihrer Entschlüsselung (vergleichbar mit Fig. 6),
• Fig. 13 das rekonstruierte Fourierspektrum (Fourierkoeffizienten gebildet aus Sprachparameter, vergleichbar mit Fig. 4, wobei die verwendete Informationsreduktion erkennbar wird),
• Fig. 14 das Fourierspektrum nach seiner inversen Gewichtung (vergleichbar mit Fig. 4),
• Fig. 15 das (vergleichbar mit Fig. 3) rekonstruierte Sprachsignal,
• Fig. 16a kodierte Sprachparametersignale über einem Intervall von 100 ms (5 Sprachabschnitte zu je 128 Abtastwerten),
• Fig. 16b die Sprachparametersignale Fig. 16a nach erfolgter Modulation,
• Fig. 16c den Effekt einer einem CCITT-Sprach-' kanal entsprechenden Bandpassfilterung (des Signals 16b) mit einem Durchlassbereich von 300 bis 3000 Hz, und
• Fig. 17 einen Qualitätsvergleich zwischen einem Originalsprachsignal Fig. 17a und einem gesichert übertragenen, rekonstruierten Sprachsignal 17b.

Show it

• 1 is a block diagram of a transmitter,
• 2 is a block diagram of a receiver in a system for secure voice transmission,
• 3 shows a sampled, pulse code-modulated speech signal of 20 ms duration,
• 4 shows the complex discrete Fourier spectrum of the signal according to FIG. 3,
• 5 shows the Fourier spectrum according to FIG. 4 according to its weighting,
• 6 shows a representation of resulting speech parameters after information reduction and triple duplication,
• 7 shows the speech parameters according to FIG. 6 after their permutation,
• 8 the speech parameters according to FIG. 7 after their coding,
• 9 shows the transmitted modulated speech parameter signal (transmitter signal f '(t)),
• 10 shows the speech signal parameters after their demodulation in the receiver (comparable to FIG. 8),
• 11 shows the speech signal parameters after their decoding (comparable to FIG. 7),
• 12 shows the speech signal parameters after their decryption (comparable to FIG. 6),
• 13 shows the reconstructed Fourier spectrum (Fourier coefficients formed from speech parameters, comparable to FIG. 4, the information reduction used being recognizable),
• 14 shows the Fourier spectrum after its inverse weighting (comparable to FIG. 4),
• 15 the reconstructed speech signal (comparable to FIG. 3),
• 16a encoded speech parameter signals over an interval of 100 ms (5 speech sections with 128 samples each),
• 16b the speech parameter signals FIG. 16a after modulation,
• 16c shows the effect of a bandpass filtering (of signal 16b) corresponding to a CCITT voice channel with a passband of 300 to 3000 Hz, and
• 17 shows a quality comparison between an original voice signal FIG. 17a and a securely transmitted, reconstructed voice signal 17b.

A transmitter according to FIG. 1 essentially has a speech parameterization unit A with a downstream data reduction stage B or 4 and a subsequent coding unit C. A receiver corresponding to the transmitter (FIG. 2) is equipped with a recoding unit C ', a parameter recovery unit B' and a speech signal generator A '.

A speech signal S (t) is first supplied in a digitization stage 1 (FIG. 1) in a known manner to a bandpass filter with a pass band from 400 to 2800 Hz. The speech signal is then digitized in a conventional manner using a 12-bit analog / digital converter at a sampling frequency of 6.4 kHz. A resulting speech section (frame), consisting of 128 data points in a time interval of 20 ms, is shown as pulse code-modulated signal s (t) in FIG. 3. In the discrete Fourier transformation stage 2 downstream of the digitization stage 1, the complex discrete spectrum of this speech section is obtained, so that a function A (f) according to FIG. 4 arises at one output. In a subsequent spectrum weighting stage 3 there is now a linear increase (pre-emphasize) for higher frequencies, so that the mean amplitude of the spectrum becomes approximately constant. With this spectrum weighting level 3 (Spectral Flattening) the dynamic range of the transmission channel (CCITT channel) is used optimally. The resulting changed energy distribution also drastically reduces the residual intelligibility of the signal and consequently increases analytical safety in a first step.

The system limits of the subsequent data reduction level B are based on the physiological properties of the human ear. Accordingly, a lower frequency band (Ib in the figures) from 400 to 1250 Hz and an upper frequency band (ub in the figures) from 1250 to 2800 Hz are provided. Again, for physiological reasons, all approximated positive amount and amount times phase-related information is transmitted in the lower frequency band, while in the upper frequency band approximated amount information is processed only via k (= 4) image samples. This information now undergoes a triple duplication in a data duplicator 5 and results in a signal A (t) according to FIG. 6, four parameters being reserved here for synchronization purposes. An encryption takes place in the subsequent data encryptor 6, see FIG. FIG. 7 a (t) with the speech parameters of the same speech section permuted to FIG. 6 (m = 1). Now coding is carried out in a known manner in a data encoder 7, the wobei times multiplied information signals being summed up alternately as a unit with alternating signs, the resulting running partial sums are formed and ℓ times appear at the output, cf. Fig. 8 A (t). A digital modulation is then carried out in a modulator 8, which is adapted to the specific bandwidth of the transmission channel, in the present case a CCITT voice channel, and after implementation in a 12-bit digital / analog converter and low-pass filtering in the form of an analog transmitter signal f ( t) (Fig. 9) is fed into this.

The transmitted speech parameter signal f '(t) (transmitted transmitter signal) reaches the input of a demodulator 9, FIG. 2; it is demodulated here so that A (t) has a profile according to FIG. 10. This signal is then fed to a synchronization detector / decoder 10 and appears at its output in the decoded form A (t) according to FIG. 11. In a downstream decoder 11, the signal curve A (t) according to FIG. 12 is achieved, which is the same Sequence of parameters as that of FIG. 6.

This signal is fed to a data converter 12, whereupon the reconstructed Fourier spectrum A (f) shown in FIG. 13 arises. In the subsequent inverse spectrum weighting stage 13, the inverse weighted Fourier spectrum is formed in accordance with FIG. 14. This spectrum A (f) undergoes a reverse transformation in the downstream discrete Fourier re-transformation stage 14 and is subsequently reconstructed into a section boundary adjuster 15 and digital / analog converter 16 (12-bit D / A converter with a low pass filter) connected in series to form a reconstruction Speech signal S '(t), Fig. 15, processed.

Furthermore, it has proven to be advantageous to connect an equalizer upstream of the receiver in FIG. 2 and to sample the input signal with the sampling frequency of 12.8 kHz, which is double the transmitter. This leaves additional data points for the detection, synchronization and reconstruction of the speech signal.

16, a representation A (t) of coded speech parameters over 5 image samples corresponding to an interval of 100 ms, further relationships can be seen. 16a shows the speech parameter signals with the arrows s symbolized synchronization information that occurs every 20 ms. The same speech parameters have a time profile according to FIG. 16b after their modulation and a profile according to FIG. 16c after bandpass filtering with a passband of 300 to 3000 Hz (which corresponds to a real transmission path).

The system shown above can be varied and further developed by the person skilled in the art within wide limits. A pseudo-random number generator 17 is already provided in the exemplary embodiment, FIG. 1, which controls the number N of samples in the speech parameterization stage. Accordingly, the speech sections in the receiver, FIG. 2, are formed by a synchronized pseudo-random generator (18).

Likewise, for example, the residual intelligibility in the transmission path can be reduced to practically zero by permutation of the speech parameters over several speech sections or through time-varying speech sections, cf. 7 with m = 1.

Fig. 17 provides proof of the quality of the system: Fig. 1 7a represents the original speech signal S (t) and Fig. 17b the reconstructed speech signal S '(t).

Further evidence of the high transmission quality could be provided by corresponding spectrograms of the speech signals S (t) and S '(t), which are not shown here for drawing reasons.

Claims (7)

1. A method for the secure transmission of speech signals on an analog transmission path which is formed by a transmitter and a receiver and which has a limited band width, pulse-code modulated speech signals being generated in the transmitter and divided into individual sections consisting of N scanning samples, subjected to orthogonal transformation, encoded and subsequently transmitted, and in the receiver, which is operated in synchronism with the transmitter, returned to a speech signal by being transformed back, characterised in that, at the transmitter side-

- in a first step of the method, the discrete complex spectrum of the respective section is calculated and amplified in such a manner that the individual spectra have a constant mean amplitude;

- this amplified speech spectrum is divided into an upper and a lower frequency band, the upper frequency band, on the one hand, being provided for items of quantity information approximated from k complex, averaged image samples and, on the other hand, the lower frequency band being provided for approximated, positive quantity- and phase-related information, and

- in a second step of the method, the resultant information signals are individually muItipIiedℓ times and, within a range of m speech sections, are exchanged in a pseudo-random manner as a unit and/or encoded in groups and are transmitted.

2. A method according to Claim 1, characterised in that a lower frequency band of 400 to 1250 Hz and an upper frequency band of 1250 to 2800 Hz are used.

3. A method according to one of Claims 1 or 2, characterised in that the positive information signals, which have been multiplied I times and exchanged in a pseudo-random manner within a range of m speech sections, are summed as a unit alternating with changing sign, the resultant current partial sums are formed, multiplied I times, matched by digital modulation to the band width of the transmission channel and, supplemented by a synchronising signal, are transmitted.

4. A method according to one of Claims 1 to 3, characterised in that the number N of the scanning samples is varied in a pseudo-random manner during the parametrisation of the speech signals. 5, A method according to one of Claims 1 to 4, characterised in that, at the input of the receiver, the received signal is first equalised and then sampled with a sampling frequency which is twice that of the transmitter.

6. A circuit arrangement for carrying out a method according to Claim 1, which, at the transmitter, consists of a series circuits of a speech parametrisation unit (A) provided with a digitising stage (1), a discrete Fourier transformation stage (2), a data reduction stage (B) and an encoding unit (C) having a data duplicator (5), data coder (7) and modulator (8) and, at the receiver, comprises a return coding unit (C') having a demodulator (9) and synchronisation detector/ decoder (10), a parameter recovery unit (B') having a data converter (12), and a speech signal former (A') having a discrete Fourier retransforma- tion stage (14), a section limit matching stage (15) and a digital/analog converter (16), characterised in that-

- at the transmitter, in the speech parametrisation unit (A), the discrete Fourier transformation stage (2) is followed by a spectrum weighting stage (3) and, in the encoding unit (C), the data duplicator (5) is followed by a data encrypting stage (6), and

- at the receiver, in the return encoding unit (C'), the synchronisation detector/decoder (10) is followed by a deciphering stage (11) and, in the parameter recovery unit (B'), the data converter (12) is followed by an inverse spectrum weighting stage (13).

7. A circuit arrangement according to Claim 6, characterised in that the speech parametrisation unit (A) is preceded in the circuit by a first pseudo-random number generator (17), the output of which controls the number N of samples, and in that the formation of sections in the receiver is controlled by a second pseudo-random generator (18) which runs in synchronism with the first such generator.

Images