EP0058318B1 - Sub-band permutation method - Google Patents

Description

The invention relates to a method for exchanging subbands of the same width of a signal frequency band according to the preamble of patent claim 1.

Such methods are known, for example from German patent specification 2,426,451, in which a circuit arrangement for obtaining interchangeable frequency grids for the harmonics of fundamental speech frequencies in devices for subband interchanging, in which on the transmission side a speech signal band is divided into subbands of equal width and by a swap of the subbands a transmission band is formed, in which the subbands contained therein can be inverted, the swap made on the receiving side being reversed, which is characterized in that the frequency at the lower limit of the voice band and half the sub-bandwidth Stand in relation to two whole numbers. This circuit arrangement has five modulators which break up the speech band into five subbands by means of suitably selected upper carrier frequencies at the outputs of the downstream unit filters. The subbands obtained in this way are converted back into the original frequency position of the voice band with the aid of a further five downstream modulators, with the subbands being interchanged by supplying the modulators with a selectable selection of five out of ten carrier frequencies.

A method for the veiled transmission of message signals has become known from German Offenlegungsschrift 2,652,607, in which the known methods of time scrambling and frequency band swapping are integrated. Here too, the effort for ring modulators, band filters, memory matrix, frequency generators and control in the analog version is quite high. In the digital version, which works according to the phase method (Weaver), only one input and one output multiplier are provided, but - because they are multiples of time - work at a high multiplication rate and their individual carriers are fed from a sine / cosine table .

US Pat. No. 2,408,692 describes a transmission system with obfuscation by converting sub-bands into the frequency range of one of the sub-bands and then converting them into length-modulated pulses which are transmitted in multiples of time.

The known bandpass methods require a relatively high level of circuitry.

The object of the present invention was therefore to specify a method for the subband swapping of the type mentioned at the outset which enables time-discrete or digital signal processing with less circuitry.

The solution is achieved with the characterizing features of patent claim 1.

The method according to the invention specifies how the frequency values for the carrier oscillations are to be selected as a function of the division into subbands and the sampling frequency, so that time-discrete or digital signal processing can be carried out in an uncomplicated manner. Both rectangular and sinusoidal carriers can be used at relatively low sampling frequencies and a simple modulator circuit. In further refinements of the invention, favorable values for concealing a CCITT voice channel are specified, which are divided into five subbands.

• Fig. 1 zeigt die Blockstruktur eines erfindungsgemäßen Sprachverschleierungssystems mittels Fünfbandvertauschung.
• Fig. 2 zeigt Spektren nach der Frequenzumsetzung mit vier Sinus-Trägerschwingungen bei zeitdiskreter Signalverarbeitung.
• Fig. 3 zeigt ein Beispiel einer Modulatorrealisierung.
• Fig. 4 zeigt Spektren nach der Frequenzumsetzung mit einer Rechteck-Trägerschwingung bei zeitkontinuierlicher Signalverarbeitung.

There now follows the description of the invention with reference to the figures:

• 1 shows the block structure of a speech concealment system according to the invention by means of five-band swapping.
• 2 shows spectra after the frequency conversion with four sine carrier oscillations with time-discrete signal processing.
• 3 shows an example of a modulator implementation.
• FIG. 4 shows spectra after the frequency conversion with a rectangular carrier wave with time-continuous signal processing.

The basic block diagram of a speech concealment system based on the five-band swap is shown in FIG. 1. The input spectrum of the speech signal Xin (f) to be concealed can be affected by undesired higher-frequency spectral components, which may have to be suppressed by an input-side low-pass filter TP 1 in order to be able to comply with the sampling theorem, for example, in the case of discrete-time signal processing with a predetermined sampling frequency fS. The resulting band-limited spectrum X (f) is divided into five sub-bands in the next step using the same band passes BP 1 to BP 5. This requires prior modulation with corresponding carrier frequencies w 1, w 2, etc. The actual scrambling takes place in the next step by means of a coupling matrix (scrambling matrix), which is controlled by a random generator, the subbands X _ν (f) _ν = 1, 2 ... 5 being randomly rearranged in short time intervals according to {i, j, k, l, m} ⊂ {1,2,3,4,5}. Then the subbands [X, (f) for v = i, j, k, I, m} are converted again with the same frequencies of the input modulators X and thus come to lie in the original frequency range of the speech spectrum X (f) in a scrambled order . Possible harmonics of the output spectrum, such as. B. unwanted modulation products or periodically recurring spectral components with time-discrete signal processing are eliminated by the low-pass filter TP 2. Y (f) is the spectrum of the disguised continuous Output signal. The system for decrypting disguised speech signals also has the block structure shown in FIG. 1, with the reception-side random generator ZG having to deliver the same random sequence as the transmitter and synchronized with it for correct decryption.

2a shows the spectrum of the speech signal (telephone quality), the control position fully representing the continuous-time signal and the additional spectral components occurring symmetrically to integer multiples of the sampling frequency fS (here the reverse position is shown below fS) with time-discrete processing.

für ν = 1 bis 5. Für eine einfache Trägerfrequenzerzeugung und für eine zeitdiskrete Signalverarbeitung ist es günstig, die Mittenfrequenzen mit

geringfügig zu modifizieren. Damit erstreckt sich das zu verwürfelnde Sprachspektrum von B bis 11 B (0,31 bis 3,41 kHz). 2 2The CCITT voice channel is defined by the limits 0.3 to 3.4 kHz, which means that the sub-bandwidth is B = 620 Hz when divided into five sub-bands. The arithmetic center frequencies of the subbands are then included

for ν = 1 to 5. For simple carrier frequency generation and for discrete-time signal processing, it is advantageous to use the center frequencies

to be modified slightly. The range of speech to be scrambled thus extends from B to 11 B (0.31 to 3.41 kHz). 2 2

für ny = 0, 1, 2 usw., v = 1 bis 5, I = s IN und i = 0, 1, ..., I - 1 als frei wählbare Parameter. Im einfachsten Fall für I = 1 und i = 0 erhält man f_ν = n_ν · B. Der Mindestwert der Rechteckträgerfrequenz darf dabei 4 B nicht unterschreiten, da sonst aufgrund der Hermitezität der Spektren reeller Signale störende spektrale Überlappungen auftreten. Diese Bedingung ergibt sich in gleicher Weise auch aufgrund des Linienspektrums der Rechteckträgerschwingung nach Fourier. Für n_ν = 0 entsprechend f_ν = 0 gilt vorgegebene Einschränkung nicht, da hierbei keine Frequenzumsetzung vorgenommen wird.For converting the subbands of the spectrum belonging to the time-continuous signal according to FIG. 2a into the same frequency band by means of rectangular carrier oscillations, the fundamental frequencies f _ν . for v = 1 to 5 to be arranged equidistant at distance B according to the formula according to patent claim 1

for ny = 0, 1, 2 etc., v = 1 to 5, I = s IN and i = 0, 1, ..., I - 1 as freely selectable parameters. In the simplest case for I = 1 and i = 0 one obtains f _ν = n _ν · B. The minimum value of the rectangular carrier frequency must not fall below 4 B, since otherwise, due to the hermeticity of the spectra of real signals, disturbing spectral overlaps occur. This condition also arises in the same way on the basis of the line spectrum of the rectangular carrier wave according to Fourier. The specified restriction does not apply to n _ν = 0 corresponding to f _ν = 0, since no frequency conversion is carried out.

The overlaps can be clearly seen in FIGS. 2b to 2d, in which a frequency conversion with f 2 = 2 B, f 3 = 3 B, f 4 = 4 B is shown, they are identified as folding products.

A folding product can also be seen in FIG. 2e, but it is not disturbing since it lies above the carrier frequency f 9 = 9 B.

und 1 mit jeweils beiden Vorzeichen sowie 0 auskommt und wobei der zu bewertende Abtastsignalwert x (kT) mittels eines Auswahlschalters, der durch eine Ansteuerlogik des Analogschalters angesteuert wird, auf den entsprechenden der beiden Eingänge eines dieser Multiplizierverstärker oder auf Nullpotenital geführt ist.3 shows a simple modulator implementation with 2 multipliers, only with the two values

and 1 manages with both signs and 0, and the sample signal value x (kT) to be evaluated is routed to the corresponding one of the two inputs of one of these multiplier amplifiers or to zero potential by means of a selection switch, which is controlled by a control logic of the analog switch.

4 shows the spectral relationships for time-continuous signal processing.

The same spectral relationships can also be achieved in the frequency range shown in FIG. 4 with a discrete-time system if the sampling frequency fS is selected appropriately. Here fS is to be determined in such a way that all harmonics of all rectangular carriers f _ν for ν = 1 to 5 come to lie symmetrically to half the sampling frequency. This ensures that due to the periodicity of discrete-time signals, no further unwanted lines are added to the lines of the rectangular beam spectrum that are present anyway. The condition for this is

jeweils für µ ein Element aus der Menge der ganzen Zahlen Z={...-1, 0, +1...}. Die Abtastfrequenz ergibt sich damit zu

In relation to half the sampling frequency, spectral lines of the sampled rectangular carrier vibrations then occur at the following frequencies:

for µ one element from the set of integers Z = {...- 1, 0, +1 ...}. The sampling frequency thus results

Expensive values for k, the associated sampling frequency fS and for n _ν are compiled in Table 1 given in claim 3, solutions in the first four lines with non-conversion of the sub-bands 3, 4 or 5 and thus saving on modulators are shown.

In addition to the reduction in effort, there is a lower sampling frequency.

oder

According to the invention, the sampling frequencies can be reduced even further if sinusoidal carrier oscillations are used for the conversion, which can be combined into corresponding sequences from only a few samples with different amounts. Relationships (3) and (4) also apply here. The following applies to the sampling frequencies

or

Because of the sampling theorem, however, k / l> 11 must be observed in order to avoid disruptive spectral overfolding. The following table 2 shows possible carrier frequencies for the values k / I = 15, 20, 24, 30 and 36 in corresponding five columns relative to the sampling frequency fS and depending on n and for favorable values of i and I:

da hierbei ein Teilbandsignal ohne Frequenzumsetzung gewonnen werden kann.Table 2 shows that in the case of k / I = 15 with 1 = 2 and i = 1, there are only four of the permissible carrier frequencies. Likewise, the last result shown there is unusable for k / l = 36 with i = 0, since in any frequency interval of width 10 B only a maximum of four of the permissible carrier frequencies can be found, so that a maximum of four subbands can be separated with the same filters. The spectra after the frequency conversion are shown for k / I = ₂₄ in FIG. 2. For the values k / l = 24 and 30, the data for the carrier frequencies which are favorable for an implementation, again relative to the sampling frequency for the individual subbands, and the center frequency of the bandpass filter are listed in Table 3 specified in claim 4. With regard to time-discrete signal processing, the most cost-effective case results from the choice of the band center frequency of the bandpass filter

since a subband signal can be obtained without frequency conversion.

Claim 5 specifies a low-effort modulator implementation by means of multipliers, with depending on the integer ratio f, / fS between 1/2 and 1/6, for 1/8, 1/10 and 1/12 only 1 to 3 absolute values for one or both Signs are required (Table 4). An exemplary embodiment for f _ν / fS = 1/8 is shown in FIG. 3 described above.

Claims (5)

1. Method for the interchanging of equally wide partial bands of a signal frequency band by means of modulators (X), behind which are connected like band pass filters (BP 1 to BP 5) as well as further modulators (X), wherein certain carrier frequencies (f 1 to f 4) are fixedly associated with the modulators (X) and a coupling matrix (scrambling matrix), which is controllable by a random generator, for the scrambling of the partial bands is provided between the band pass filters (BP 1 to BP 5) and the modulators (X) connected therebehind, characterised thereby, that the carrier frequencies display the values

with v=1, 2 ... partial bands of the bandwith Band i, I and nν ε |N=1, 2 ... (elements ε from the mass of natural numbers) as freely selectable parameters, wherein i and n, can also assume the value 0, and that the modulators (X) operate discretely in time and the relationship

applies for the scanning frequency of the carrier ascillations.

2. Method according to claim 1, characterised thereby, that the carriers are sinusoidal and the scanning frequency is chosen as

with KGV equal to the lowest common multiple or as

3. Method according to claim 1, wherein the signal frequency band is the CCITT speech channel (0.3 to 3.4 Kilohertz), characterised thereby, that n=5 partial bands of the bandwidth B=620 Hertz are formed and that the frequency translation takes place with rectangular carriers according to one of the following possibilities of the table 1:

4. Method according to claim 1 or 2, characterised thereby, that n = 5 partial bands of the bandwidth B = 620 Hertz are formed and that the frequency translation takes place with sinusoidal carriers according to one of the following possibilities of the table 3:

5. Method according to claim 2 or 4, characterised thereby, that the modulators (X) consist of a multiplier circuit which in the rhythm of the scanning frequency fS multiply the signal, for fy/fS or integral multiple thereof, according to the following table 4 by the factors stated there:

Images