RU2120700C1 - Method for voice signal processing and device which implements said method - Google Patents

Abstract

FIELD: communications. SUBSTANCE: time-base encrypting method involves splitting voice signal into sequence of time intervals, splitting resulted intervals into sequence of sub-intervals, their random permutation. Duration of said intervals is, number and duration of sub-intervals are random. Direction for reading signal from sub-intervals produced by splitting is random. Sub-intervals which are randomly selected is subjected to reading of converted signal with speed is greater than speed of their storing. Pauses caused by this difference in speed are located in time interval in random position and are filled with background noise which characteristics depend on current pseudorandom sequence. Corresponding device has unit for interface to channel of voice signal to be transmitted, which is connected to first band-pass filter, unit for interface to channel of voice signal to be received, which is connected to second band-pass filter. Outputs of first and second band-pass filters are connected to input of analog-to-digital converter. In addition device has digital-to-analog converter which output is connected to second commutator, which one output is connected to third band-pass filter which output is connected to input of unit for interface to output encoded voice signal. Second output of second commutator is connected to fourth band-pass filter, which output is connected to unit for interface to output decoded voice signal. In addition device has processor which serves as unit for synchronization of communication between users and pseudorandom sequence generator which restores spectrum of noise signal, performs pseudorandom processing of sequences of digital samples of voice signal, controls device operations and runs self tests. EFFECT: decreased probability of surveillance decoding, simplified design, increased functional capabilities.

Description

 The invention relates to communication technology, to methods for encoding speech signals.

 A known method of encoding speech signals using time encryption, which consists in dividing the speech signal into time intervals, dividing the received intervals into a number of sub-intervals and rearranging them according to a random law / 1 // 2 /. A device that implements this method contains the same number of storage devices (memory) for the encoder and decoder, memory have a total capacity numerically equal to the number of samples over the time interval of samples; the device also contains selector switches that select a particular memory in accordance with the law, equally formed by the pseudo-random sequence sensors in the encoder and decoder, thereby randomly alternating time subintervals within the sampling intervals is achieved; the device includes counters for counting the number of samples within time intervals, counters for counting the number of samples within randomly selected sub-intervals, a counter for counting the number of sub-intervals in the sample interval; the device contains a clock generator, a pseudo-random sequence sensor, a circuit that controls the synchronization process of subscribers and controls the synchronism of the components of the device, filtering and processing of input and output analog signals, an analog-to-digital converter, and a digital-to-analog converter.

 A known method of encoding speech signals using time encryption, which consists in dividing the speech signal into time intervals, dividing the received intervals into a number of sub-intervals and rearranging them according to a random law, then the sub-intervals are subjected to temporary compression or temporary expansion in accordance with the random law / 6 // 7 /. A device that implements this method includes a preliminary filtering circuit of the incoming speech signal, an analog-to-digital converter, a delay line for a double length of the sampling interval, an addressing scheme of the memory elements of the delay line in accordance with the law generated by the pseudo-random sequence sensor, counters for counting the number of samples in the interval transformations in the encoder and decoder, a clock generator, a circuit for controlling the process of synchronizing subscribers, a circuit for writing to and reading from the memory, digital-to-analogue th converter, filter circuit of the output speech signal.

 The closest in technical essence to the claimed method and device is / 2 /.

 Method / 2 / consists in dividing the speech signal into time intervals, dividing the obtained intervals into a number of sub-intervals and rearranging them according to a random law.

 The disadvantage of this method is the low degree of security of speech signals from decoding by an outside observer.

 The objective of the invention is to reduce the likelihood of decoding speech signals by an outside observer.

The problem is solved in that in the known method, which consists in dividing the speech signal into time intervals, dividing the obtained intervals into a number of sub-intervals and permuting them according to a random law, the following are introduced:
- selection of the duration of time intervals according to a random law;
- the choice of the number and duration of time sub-intervals in the sampling interval according to a random law;
- for the resulting sub-intervals, the direction of reading the signal is selected according to a random law;
- on the subintervals selected according to a random law, the converted signal is read at a speed exceeding the recording speed, and the temporary pauses resulting from the difference in speeds are randomly located on the time interval and filled with a noise signal, the parameters of which depend on the type of the current pseudo-random sequence.

 The device that implements the method / 2 / contains the same number of memories for the encoder and decoder, selector switches for selecting the memory, pseudo-random sequence sensor, counters for counting the number of samples within time intervals, counters for counting the number of samples inside randomly selected sub-intervals, counters for counting the number of sub-intervals in the sampling interval, a clock generator, a pseudo-random sequence sensor, a circuit that controls the synchronization process subscriber reduction and synchronization control of the operation of the device components, filtering and processing schemes of input and output analog signals, analog-to-digital converter, digital-to-analog converter.

 The disadvantage of the device that implements the method / 2 / is its complexity and cumbersome.

 The objective of the invention is to simplify the hardware implementation of the device while increasing its functions to implement the proposed method of encoding speech signals.

 The problem is solved in that in a known device containing a device for interfacing with a channel for an outgoing speech signal connected to a band-pass filter; a channel interface device for an input coded speech signal connected to a bandpass filter; the bandpass filter outputs are connected to the inputs of the switch, the output of the switch is connected to the input of an analog-to-digital converter; a digital-to-analog converter, the output of which is connected to the switch, one output of the switch is connected to a band-pass filter, the output of which is connected to a coupler for the output of the encoded speech signal, the other output of the switch is connected to a band-pass filter, the output of which is connected to the coupler for the output of the decoded speech signal, a processor programmatically performing the functions of a subscriber synchronization device and a pseudo-random sequence generator, performing restoration dividing the spectrum of the noise signal, pseudo-random conversion over sequences of digital samples of the speech signal to reduce the likelihood of decoding the signal by an outside observer, component management and device control.

 The speech signal is sampled on a time interval of pseudorandom duration T (Fig. 1), obtained by digital samples, fill the buffer, pseudo-random choice of duration and number of possible time partitions t on time interval T is performed by calculating the starting and ending addresses of the sequences of samples stored in the buffer, pseudo-random permutation of a series of samples by specifying the order of their reading, pseudo-random selection of the reading direction for each the sequence of samples, the restoration of the analog waveform at pseudo-random time intervals is performed at a frequency that exceeds the recording speed, the resulting time pauses are pseudo-randomly placed on the interval T and filled with noise signal S (Fig. 2), the parameters of which depend on the type of the current pseudo-random sequence. The coding efficiency of the speech signal increases from the application of the totality of the proposed coding methods.

 Simplification of the hardware implementation is achieved by introducing a processor that performs the described conversion programmatically, synchronizes subscriber devices, generates a pseudo-random sequence (PSP), and monitors linear circuits and memory devices.

 In FIG. 3 shows a structural diagram of a device for implementing the method.

 The device comprises a device for interfacing with circuits for passing an outgoing speech signal (US1) 1, a device for interfacing with circuits for passing an encoded speech signal (US2) 2 coming from a channel, a first band-pass filter (PF) 3, a second band-pass filter PF 4, and a first switch 5 , analog-to-digital converter (ADC) 6, a processor 7 containing a microprocessor (MP) 8, and random access memory (RAM) 9 and read-only memory (ROM) 10, digital-to-analog converter (DAC) 11, second switch 12, third floor owl filter PD 13, PD fourth bandpass filter 14, interface unit with the output circuits to issue a coded speech channel (HSS) 15, an interface with the output circuits to issue subscriber decoded speech signal (US4) 16.

 The output of US1 1 is connected to the input of PF 3, the output of US2 2 is connected to the input of PF 4, the output of PF 3 is connected to the first output of switch 5, the output of PF 4 is connected to the second input of switch 5, the output of switch 5 is connected to the input of ADC 6, the outputs of ADC 6 connected to MP 8, RAM 9 connected to MP 8, ROM 10 connected to MP 8, DAC 11 inputs connected to MP 8, DAC output 11 connected to switch 12, first output of switch 12 connected to PF 13, second output of switch 12 connected to PF 14, the output of PF 13 is connected to US3 15, the output of PF 4 is connected to US4 16.

 The device operates as follows.

 US1, US2, US3, US4 are used to connect the device to the communication channel, the spectrum of the outgoing speech signal is limited from above to below by the PF 3, through the switch 5 it goes to the ADC 6, where it is sampled under the control of the processor 7, the selected number of samples is placed in the RAM buffer 9 , MP 8 digitally performs the described conversion according to the program recorded in ROM 10, and in the specified order reads samples from RAM 9 to DAC 11, from the output of DAC 11, the encoded speech signal passes through switch 12 and is filtered by PF 13.

 Above the incoming encoded speech signal going through the PF 4 circuits, switch 5, ADC 6, RAM 9, DAC 11, switch 12 and PF 14, the processor performs the inverse transformation according to the program recorded in ROM 8.

 The functions of the subscriber synchronization device are the generation and reception of the trigger combination along the speech signal paths, the generation of a pseudo-random sequence of a given length, the pseudo-random conversion of a sequence of digital samples stored in RAM, the control of linear circuits and memory, service functions are performed under the control of the processor using a program written in ROM. The gain is achieved as a result of the software implementation of the above functions, which does not lead to an increase in weight and size characteristics and does not require large additional resources - processor performance, the amount of memory used.

 In FIG. 4 shows a functional diagram of a device that implements the described method. The diagram shows the central processor (CPU) 1, a clock pulse generator with a frequency division circuit (GTI) 2, a device for interfacing with circuits of a voice signal coming from a subscriber (CSS1) 3, a device for interfacing with circuits of passing a coded speech signal (CSS2 coming from a channel) ) 4, a band-pass filter (PF) 5, PF 6, address register (РГА) 7, address selector (SA) 8, analog key (KA) 9, RAM 10, ADC 11, an operational amplifier with a given gain and offset to the region ADC conversion (OU) 12, ROM 13, DAC 14, KA 15, PF 16, PF 17, m trunk transceiver (MPP) 18, control register (RGU) 19, a device for interfacing with output circuits for issuing a coded speech signal (US3) 20 to a channel, a device for interfacing with output circuits for issuing a decoded speech signal (US4) 21 to a subscriber, front panel 22 on which the keyboard matrix 23 and indicators 24 are located.

 The original speech signal is fed to US1 3; the output of US1 3 is connected to the input of the PF 5; the output of the PF 5 is connected to the first input of the spacecraft 9; the incoming coded speech signal is supplied to US2 4; the output of US2 4 is connected to the input of PF6; the output of US2 4 is connected to the second input of the spacecraft 9; the output of the spacecraft 9 is connected to the input of the OS 12; the output of the OS 12 is connected to the input of the ADC 11; the output of the DAC 14 is connected to the KA 15; the first output of the spacecraft 15 is connected to the input of the PF 16; the output of the PF 16 is connected to the input of US3 20; output US3 20 remove the output encoded speech signal; the second output of the spacecraft 15 is connected to the input of the PF 17; the output of the PF 17 is connected to the input of US4 21; output US4 21 remove the output decoded speech signal; CPU 1 is connected via a multiplexed address-data highway with RgA 7, with RAM 10, with ROM 13, with MPP 18, with ADC 11, with DAC 14, with RgU 19; with RGA 7, the address lines are connected to RAM 10, to ROM 13, to CA 8; with CA 8, the device selection lines are connected to RAM 10, to ROM 13, to MPP 18, to ADC 11, to DAC 14, to RGU 19; from the matrix of the keyboard 23 lines of analysis of the keys are connected to the inputs of the MPP 18; with РГУ 19 keyboard service line connected to the keyboard matrix 23; with РГУ 19 indicator control lines are connected to indicators 24; with RGU 19 the signal line of the start of conversion conversion is connected to the ADC 11; with РГУ 19 the input key control line is connected to KA 9; with RGU 19 the output key control line is connected to KA 15; with RGU 19, the control line connecting the device to the channel is connected with US1 3, with US2 4, with US3 20, with US4 21; with ADC 11, the conversion end line is connected to MPP 18; with GTI 2, the frequency sampling signal line Fд is connected to CPU 1; with the GTI 2, the signal line of the clock frequency F1 is connected to the CPU 1; with GTI 2 line signal conversion frequency F2 is connected to the ADC 11.

 The operation of the device is carried out under the control of CPU 1 according to the program located in ROM 13. A clock signal with a frequency of F1 is supplied to CPU 1 from the GTI 2. A cycle of a single exchange on the address-data line between CPU 1 and devices for performing data read-write operations begins setting CPU 1 to the device address highway. The device address is stored on РГА 7, it is held by it on the address inputs of RAM 10 and ROM 13, and also goes to CA 8. CA 8 selects the addressable CPU 1 of the device (RAM 10, ADC 11, ROM 13, DAC 14, MPP 18, RGU 19).

With a frequency equal to the sampling frequency Fd, CPU 1, following the commands selected from ROM 13, executes the algorithm cycle in the mode - with conversion, which consists in the following:
The CPU 1 switches using the CA key 9 KA 9 so that it switches the PF 5 with the OS 12, thereby supplying the input of the voice signal from the subscriber through the US1 3, PF 5, KA 9, OU 12 to the ADC 11 input;
The CPU 1 through RgU 19 provides the ADC 11 signal to start the conversion; by a signal from the ADC 11 - the conversion is done, read through the MPP 18, the CPU 1 reads a digital sample from the ADC 11 and places it in the calculated RAM cell 10, which is included in the buffer for the signal from the subscriber;
CPU 1 switches KA 9 with the help of PrGU 19 so that it switches PF 6 with OA 12, thereby supplying the encoded speech signal coming from the channel to the ADC 11 input, which has passed through US2, PF 6, KA 9, OU 12 circuits;
The CPU 1 through RgU 19 provides the ADC 11 signal to start the conversion; by a signal from the ADC 11 — the conversion is done, read through the MPP 18, the CPU 1 reads the digital sample from the ADC 11 and places it in the calculated RAM cell 10, which is included in the buffer for the encoded speech signal coming from the channel;
The CPU 1 switches using the RCU 19 key KA 15 so that it commutes the output of the DAC 14 with the PF 16, thereby forming a circuit for the output of the encoded speech signal - DAC 14, KA 15, PF 16, US3 20; CPU 1 reads a speech signal sample from a calculated RAM buffer cell depending on the encoding parameters determined by the type of the current pseudo-random sequence, and provides a sample to DAC 14, from the output of which an analog encoded speech signal is filtered on PF 16 and transmitted through the US3 20 to the communication channel;
The CPU 1 overlaps the key KA 15 with the help of РГУ 19 so that it switches the output of the DAC 14 with the PF 17, thereby forming a circuit for the output decoded speech signal - DAC 14, KA 15, PF 17, US4 21; CPU 1 reads a speech signal sample from a calculated RAM buffer cell depending on the decoding parameters of the determined types of the current pseudo-random sequence and provides a sample to DAC 14, from the output of which an analog decoded speech signal is filtered on PF 17 and issued to the subscriber via US4 21.

 From the sequence of current samples for the incoming encoded speech signal, the CPU 1 tries to extract a sync sequence (a sync sequence of a certain kind is stored in ROM in the form of a mask). If a sync sequence is found, CPU1 synchronizes its work on it.

 After a specified time, determined by the allowable amount of desynchronization, or when connected to the channel, the CPU 1 generates a sync sequence of a certain type, which goes along the paths of the output encoded speech signal. CPU1 using RGU 19 and MPP 18 polls the keys of the matrix 23 of the front panel 22 and serves the indicators 24.

 One cycle of the algorithm passes for a quantum whose duration is inversely proportional to the sampling frequency.

 The CPU 1 once during time T, separately for the outgoing and incoming speech signals, generates a pseudo-random number from a sequence of a given length and, depending on its type, calculates the encoding parameters for the signal coming from the subscriber and the decoding parameters for the signal incoming from the channel, and also makes a choice parameters for the synthesis of a speech-like noise signal.

 The conversion time of the ADC 11 is proportional to the frequency F2. A signal of this frequency is fed to the ADC 11 with the GTI 2.

 Enter the key that determines the type of pseudo-random sequence and set the device operating modes (with and without conversion), display input and output information about the device status, service functions are carried out using the front panel 22 - keyboard matrix 23 and indicators 24 located on it and served by software under the control of the CPU 1. Entering information from the front panel (pressing a key) is accompanied by the issuance of a short-term tone signal to the subscriber along the speech signal circuits, reg echoed by him and is a response to a keystroke. The tone transmission of a given duration and frequency is performed under the control of the processor.

 Separate alternate switching on / off of the coding (transmitting) and decoding (receiving) parts of the device through the pairing devices for channels with half-duplex communication and simplex communication is allowed. The transmitting part includes US1 3 and US3 20, the receiving part includes US2 4 and US4 21.

 When the device is turned off from the transmission channel from the front panel 22, the device remains in a state connected to the channel for a time equal to T, then a software disconnection of the interface devices for transmission is implemented. This ensures that messages are transmitted without loss due to conversion delay times.

 The operation of the device in the mode - without conversion allows you to control linear circuits, processor and software. The signals in this case are transmitted without delay to the RAM from input to output under the control of the processor with a frequency equal to the sampling frequency (a “transparent” channel is implemented - without encoding). The analysis of the health of the component parts of the device is performed by the processor with the issuance of control results to the subscriber. In this case, the serviceability of the device is registered by the subscriber visually by the indicators of the front panel, serviced by the processor, and by ear with the assessment of intelligibility.

The conversion parameters are selected taking into account the following conditions:
- the speech signal in the channel undergoes a delay for a total time equal to 2T;
- the time interval T is limited from above by the maximum allowable delay in bringing the speech signal to the subscriber;
- the time interval t must not exceed the time interval T;
- time intervals T and t are limited from below by the effect on intelligibility of the possible allowable accuracy of synchronization;
- the expansion of the spectrum band due to the difference in the write-read speeds of the samples should not lead to the exit of the used channel bandwidth;
- the number of quantization levels is selected depending on the requirements for speech intelligibility;
- the sampling rate is selected taking into account the processor performance, used RAM capacity.

 Using the proposed encoding method allows to obtain encrypted speech signals with a high degree of security from decoding by an outside observer, making unauthorized listening to confidential conversations less likely.

 The proposed hardware solution effectively implements the described encoding method.

In a specific implementation of the proposed method, the following parameters were used:
- the duration of the interval T is selected in a pseudo-random manner in the range of 800 - 1300 ms with a step of 20 ms;
- the number of sub-intervals t is selected in a pseudo-random manner from 1 to 6;
- the duration of the sub-intervals t is selected in a pseudo-random manner and not less than 200 ms;
- pseudo-random choice of the reading order of the subintervals t;
- pseudo-random choice of the reading direction of the subintervals t;
- pseudo-random selection of subintervals t at which reading is performed at a speed exceeding the write speed, the resulting temporary pauses according to a random law are located on the interval T and are filled with a noise signal, the characteristics of which depend on the type of the current memory bandwidth;
- sampling frequency 8 kHz;
- the number of quantization levels 256 (count-byte-8 bits);
- PF cut-off frequencies by level - 6 dB 300 Hz - 1.8 (3.5) kHz;
- sync sequence length 13 bits, type of signal TH;
- accuracy of synchronization +/- 1 ms;
- allowable desynchronization +/- 5 ms for 3 minutes;
- length of memory bandwidth 17 bits;
- processor performance 1.4 million op / s;
- used RAM volume is 32 kb;
- The used volume of ROM is 2 KB;
- ADC conversion time 5 μs.

Claims (1)

 A method for encoding speech signals by means of time encryption, which consists in dividing the speech signal into time intervals, dividing the obtained time intervals into a number of sub-intervals and rearranging them according to a random law, characterized in that when dividing the speech signal, the duration of the time intervals is formed according to a random law, according to which in each time interval, the number and duration of the sub-intervals and the direction of reading the speech signal in the sub-intervals are determined, according to a random law subintervals are read, the speech signal is read at a speed exceeding the recording speed, the resulting time pauses are placed on the time interval between the subintervals according to a random law and filled with a noise signal, the parameters of which are formed according to a random law.

Images