JPS624021B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a confidential transmission device for audio signals using spectrum swapping, and particularly to a confidential transmission device suitable for communications requiring high privacy. Conventional confidential message transmission devices can be roughly divided into two types: devices that scramble on the time axis and devices that scramble on the frequency axis. The former devices include (i) devices that temporally divide the audio signal into blocks and change the order of sample values within each block, and (ii) devices that change the polarity of sample values. As belonging to the equipment, (iii)...device using frequency inversion method, (iv)...device that divides the audio signal into a plurality of frequency slots on the frequency axis and replaces these slots according to a predetermined rule. There are also devices that invert the frequency within one slot, and (v) devices that temporally change the switching rule in (iv) above. Criteria for evaluating the characteristics of a privacy device include diversity in the selection of encryption keys and decryption keys, confidentiality when the encryption key and decryption key do not correspond, and whether the decryption key is correct for the encryption key. The quality of the audio when it is selected and decoded, the scale of the device, etc. can be considered. In the case of (i) and (ii) of the conventional device described above, if the frequency of order shuffling and polarity reversal is increased to improve confidentiality, the scrambled output signal becomes white noise and the transmission path The disadvantage is that the intelligibility of the descrambled voice deteriorates due to the frequency characteristics of the descrambled voice. In addition, the frequency inversion method (iii) is easy to decode, and the device (iv) requires a narrow bandwidth of the frequency slot to improve confidentiality, so a filter with sharp frequency characteristics is used. Since a large number of them are required, the disadvantage is that the scale of the circuit becomes large. Furthermore, in a device like (v) that changes slot exchange over time, decoding is more difficult than in (iv), but because the temporal response of the narrowband filter is slow, the descrambled output signal However, there is a drawback that noise is large due to switching of the slot switching order, that is, switching of the narrow band filter. The present invention has been made in view of the drawbacks of the prior art described above, and is based on the concept (v) above. After applying the Hadamard transform (FHT) and transposing the spectrum, the inverse transform is performed and transmitted, and the receiving side performs FFT or FHT again to restore the transposing on the transmitting side, and then performs the inverse transform of FFT or FHT. The purpose of the present invention is to provide a voice signal secrecy device in which a decoded signal is obtained by decoding the voice signal. In this device, the number of spectra increases by increasing the number of FFT (or FHT) points, so it is easier to increase the number of keys for encryption and decryption compared to conventional technology. It can make deciphering even more difficult and increase confidentiality. Furthermore, in the case of conventional technology using narrowband filters, slot swapping was performed using a device that uses the same number of modulators as the number of slots, but in the present invention, since sample values are handled, swapping can be easily performed using arithmetic processing. There is an effect that can be done. Furthermore, since a narrow band filter with a blunt impulse response is not used, there is an advantage that noise due to switching of the order of spectrum switching can be reduced in the descrambled output signal. The present invention will be explained in detail below using the drawings. FIG. 1 shows an embodiment in which fast Fourier transform (FHT) is applied as orthogonal transform to an audio signal. In the figure, 1 is the input terminal, 2 is the FFT
(Fast Fourier Transform) circuit, 3 is a spectrum switching circuit, 4 is an IFFT (Inverse Fast Fourier Transform) circuit,
5 is a synthesis circuit, 6 is a spectrum switching control circuit, 7
Reference numeral 8 indicates a timing generation circuit, 8 indicates a synchronization signal generation circuit, and 9 indicates an output terminal. An audio signal in the 4KHz band inputted from the input terminal 1 is converted into a signal on the frequency axis by the FT circuit 2 to generate a spectrum. This spectrum is swapped by the spectrum swapping circuit 3 according to a predetermined rule instructed by the spectrum swapping control circuit 6, and then converted back into a time series signal by the IFFT circuit 4. It is combined with the synchronization signal and output to the output terminal 9 as a scrambled output signal in the 4KHz band. The above operation is performed by the timing generation circuit 7.
controlled by. Also, in this example
The FFT circuit 2 and the IFFT circuit 4 may use the same FFT circuit in a time-sharing manner. Figure 2 shows the spectrum switching circuit 3 shown in Figure 1.
A configuration example of the spectrum switching control circuit 6 will be shown, and the spectrum switching operation will be explained in detail. In Fig. 2, 10 is the FFT circuit 2 of Fig. 1.
remembers the output (spectrum) from
RAM, 11 is ROM that stores the exchange key, 1
2 remembers the swapped spectrum
RAM, 13 is a counter indicating the replacement key number,
14 indicates a counter indicating the address of the ROM 11. Furthermore, the three parts divided by broken lines or solid lines are the timing generation circuit 7 in FIG.
It corresponds to the switching control circuit 6 and the spectrum switching circuit 3, and data in the RAM 12 is output to the IFFT circuit 4. Table 1 is used to assist in explaining the operation in Figure 2, and the contents of the ROM 11 in Figure 2 are as follows:
The diagram is shown assuming that there are eight spectra and four types of exchange keys. In Table 1, the address is represented by 5 pits, the upper 2 bits correspond to the key number, and the lower 3 bits correspond to the 0th to 7th spectra. For example, in the 0th decimal key, each spectrum is scrambled 4→
0, 3 → 1, 5 → 2, 2 → 3, 6 → 4, 1 → 5,
7 → 6, 0 → 7, and by descrambling, 7 → 0, 5 → 1, 3 → 2, 1 →
3, 0 → 4, 2 → 5, 4 → 6, 6 → 7, etc., respectively, indicating that they are exchanged and returned to their original state. The operation of the circuit shown in FIG. 2 will now be described. Set the number of the first key to be used to 13. For example, assume that the content of 13 is "10". The upper two bits of the counter 14 are set to the same bit pattern as the counter 13, and the lower three bits are reset to "000" before starting operation. These 5 bits (10000) become the address of the ROM 11. As can be seen from Table 1, the value read from the ROM 11 is 5 (101), which becomes the address of the RAM 10. The value to be read out (the value of the fifth spectrum) is written into the RAM 12 using the lower three bits (000) of the counter 14 as an address. Next, a timing pulse is output from the timing pulse generator 7, and "1" is added to the counter 14 to become "10001", and the same operation as above is repeated. The replacement of 8 spectra is completed.

【table】

[Table] Then, the counter 14 is reset to "10000". When exchanging several frames with the same key, "1" is added to the counter 13 and becomes "11".
Replacement will be performed using the key with decimal display "3". Note that when performing FFT on N points, one frame is composed of N sample values. If a replacement operation is performed using the key with decimal representation "3" over several frames, the counter 13 will be reset to "00" and the key with decimal representation "0" will be used. By repeating the above operations, it becomes possible to scramble a highly confidential audio signal. Next, an example of synchronization between reception and transmission (frame synchronization) when receiving a signal sent out according to the present invention will be described. There are several possible methods of notifying the receiving side of the synchronization information, but here we will use a method of inputting a pilot signal outside the audio band. ₁ in Figure 3
is the voice band lower limit frequency, ₂ is the upper limit frequency, _c
is the synchronization signal carrier frequency, ₃ is the transmission line band upper limit frequency, and _s is the sampling frequency.As shown in the figure, the carrier wave with frequency _c is the frame frequency _r [ _r = _s / N (N...number of FFT points) A method is used in which a signal modulated with a sine wave of about several tens of Hz is placed within _two to _three bands. This signal is generated by the synchronization signal generating circuit 8 and combined with the scramble signal by the combining circuit 5. FIG. 4 shows an example of a circuit on the receiving side for receiving signals transmitted by the apparatus of the present invention. In the figure, 15 is an input terminal, 16 is a synchronization signal removal filter, 17 is a synchronization signal extraction circuit,
Reference numeral 18 indicates an output terminal, from which a scrambled signal is input, and from 18 a decoded audio signal is output. The other parts 2a, 4a, 6a, and 7a are almost the same as 2, 4, 6, and 7 of the embodiment shown in FIG. However, the difference between this embodiment and the transmitting embodiment shown in FIG. 2 is that the data in the rightmost column of Table 1 is written in the ROM 11 provided in the spectrum switching circuit 3. That's true. Therefore, detailed explanation will be omitted. The following points can be considered regarding spectrum swapping. (b) As shown in Figure 3, the transmission line band is ₁ ~
₃ , when ₂ to ₃ are required as the synchronization signal band, the important band for the audio signal is ₁ to 3.
It is necessary to replace it so that it fits within ₂ . (b) Let the sampling frequency of the audio signal be _s and the sample value be _Vi (i=0, 1,...), then N
Spectrum S _k obtained when point FFT is applied
teeth However, the following formula As can be seen from, S _k (k=0, ..., N/2)
corresponds to frequencies from 0 to _s /2,

[Formula] (k=1,...,N/2-1) corresponds to negative frequencies and is a complex conjugate with S〓 _-k . Therefore, the calculation will be simplified as follows. That is, the values of only the spectra of k=0 to N/2-1 are determined by FFT and stored in the RAM 10 in FIG.
This is subjected to spectrum swapping and stored in the RAM 12, but for the spectra from k=N/2+1 to N-1, the corresponding positive frequency spectra with their imaginary parts inverted in sign are stored in the RAM 12. Use as input. FIG. 5 shows an embodiment in which FIG. 2 is modified to simplify calculations. In Figure 5, 19 is an address conversion ROM, 20 is an AND circuit, and 21 is an AND circuit.
This circuit inverts the sign of data only when the output of the circuit 20 is "1". Also, Tables 2-1 to 2-4
is used as an aid to the explanation of the operation in FIG. 5, and shows the contents of RAM and ROM. In this embodiment, the number of FFT points is 8 (N=8), the ROM 11 stores four types of exchange keys (Table 2-1), and the RAM 10 stores K=0 to N/ The spectrum of 2-1 is stored (Table 2-2). Table 2-3 shows the contents of the RAM 12 when it is assumed that scrambling is performed using the key with decimal representation "1" in Table 2-1. Table 2-4 shows the conversion of the lower 4 bits of the 14 counters.
The contents of the address conversion ROM 19 are shown to be used as addresses of the RAM 12.

【table】

【table】

【table】

【table】

[Table] The operation of the circuit shown in FIG. 5 will be explained next. Table 2
When using the decimal representation "1" key of -1, the counter 14 starts from "010000", but considering the process currently in progress, assume that the counter 14 is "011000". Using these upper 4 bits “0110” as an address, the data read from ROM11 is “11”
The address of RAM10 is “110” and Re
(S ₃ ) is output as data. This data
It is written to RAM 12, and the address starts from the lower 4 bits of counter 14.
The address converted by the ROM 19 (ie, "0100") is used. Next, "1" is added to the counter 14 to become "011001", but the addresses of ROM 11 and RAM 10 remain unchanged, and the address of ROM 19 becomes "1001". Therefore, Re(S ₃ ) is stored at address “1100” in the RAM 12. This has the role of storing the real part of the spectrum of negative frequencies. When the counter 14 is added to “011010”, the RAM 10 address becomes “111” and Im
(S ₃ ) is output. This is stored at address "0101" in RAM 12 (ie, output at address "1010" in ROM 19). Next, when "1" is added to the counter 14 and becomes "011011",
Data will be stored at address "1101" in RAM 12. In this case, AND circuit 20
The output becomes "1" and the sign of the data is inverted by the sign inverter 21, so that -Im(S ₃ ) is stored. Next, when "1" is added to the counter 14, it becomes "011100", the address of the ROM 11 becomes "0111", the next spectrum exchange is performed, and the above steps are repeated. In addition, in Table 2-4, the data content is
However, this is “0000”, “0001”,
Indicates that any appropriate value other than "1000" or "1001" may be used. This location corresponds to the negative frequency component for K=0 and does not require writing, but is stored at an appropriate address in the RAM 12 to simplify the circuit. (There is no problem because the correct value will be written to this stored address later.) By the above spectrum swapping, the spectrum value is stored in the RAM 12 addresses “0000” to “0111”, “1010”, and “1111”. However, since "1000" and "1001" represent the spectrum of K=N/2, they are cleared to 0 at the time of reset so that they are not written. In the case of FFT, a spectrum is obtained when a time-series signal within one frame is repeated infinitely, so it does not perfectly match the actual frequency. Therefore, distortion occurs when descrambling is performed after passing through the transmission path, but if the number of FFT points is set to several tens or more, the effect is small and there is no problem. Next, a case will be described in which a fast Hadamard transform (FHT) is applied as an orthogonal transform to an audio signal. The circuit configuration is exactly the same as in the case of FFT, only that the FFT circuit 2 in FIG. 1 is replaced with an FHT (fast Hadamard transform) circuit, and the IFFT circuit 4 is replaced with an IFHT (inverse fast Hadamard transform) circuit. The spectrum is also slightly different from that of FFT. That is, when the sampling frequency of the audio signal is _s and the N-point Walsh type FHT is applied ( ₀ = _s / N), in the resulting spectrum S _k (k = 0, ..., N-1), S ₀ is the magnitude of the DC component, S ₁ is the sine wave component of frequency ₀ , S ₂ is the cosine wave component of frequency ₀ , S ₃ is the sine wave component of frequency ₂₀ , S ₄ is the cosine wave component of frequency ₂₀ ,...
..., S _N-1 corresponds to a sine wave component with a frequency of N/ _2.0 , and differs from FFT in that it is a real value. When comparing FFT and FHT, the following two points can be considered as characteristics. (a) Regarding the calculation method, it consists only of addition and subtraction.
FHT is easier. (b) Regarding the spectrum, FFT is closer to the actual frequency, so the sound quality during descrambling is better. As described above, the device of the present invention performs fast Fourier transform or fast Hadamard transform on the audio signal, scrambles it, and sends it out to the transmission path. It is clear that the system is superior in terms of audio quality and scale of equipment.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a transmitting side embodiment of the present invention, FIGS. 2 and 5 are block diagrams showing specific examples of circuits used in the present invention, and FIG. 3 is for explaining a transmission signal according to the present invention. FIG. 4 is a block diagram showing an example of a receiving side circuit for receiving signals transmitted according to the present invention.

Claims (1)

[Claims] 1. A first FFT means that performs fast Fourier transform on an audio signal to be transmitted to create a spectral array, and permutes the created spectral array according to a predetermined rule. a first switching means for obtaining a scrambled output; a first IFFT means for transmitting a signal obtained by subjecting the scrambled output to an inverse fast Fourier transform to a transmission path; a second FFT means for generating a spectral array by performing a spectral array; and a second transposition for obtaining a scrambled output by re-commuting the created spectral array to restore the transposition according to the predetermined rule. and second IFFT means for performing inverse fast Fourier transform on the scrambled output to obtain a decoded output. 2. A first FHT means for creating a spectral array by subjecting the audio signal to be transmitted to a fast Hadamard transform, and a first means for obtaining a scrambled output by permuting the created spectral array according to a predetermined rule. a first IFHT means for transmitting a signal obtained by subjecting the scrambled output to an inverse fast Hadamard transform to a transmission path, and creating a spectral array by performing a fast Hadamard transform to the signal transmitted by the transmission path. a second shuffling means for obtaining a scrambled output by reshuffling the created spectrum array to restore shuffling according to the predetermined rule; and a second shuffling means for obtaining a scrambled output. and a second IFHT means for obtaining a decoded output by performing inverse fast Hadamard transform on the voice signal.