JP2013098869A - Voice system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve a conventional problem that, particularly in analog communication, a voice waveform is deformed using various methods to increase communication secrecy, and however, once voice is understood partially, an original input voice may be restored from this clue, because the deformation is limited to a simple replacement of an input voice waveform block.SOLUTION: An element piece of the input voice is replaced with another element piece using a cipher key and transmitted. By this, restoration by the understanding of a partial voice is not possible, so that the communication secrecy is improved.

Description

  The present invention relates to a technique for improving secrecy in communication, particularly analog communication.

Communication concealment is an essential technology in modern times. On the other hand, a communication path has been established for a long time in analog voice communication among communication methods, and communication can be easily performed by using this. However, analog communication has a problem in that it is vulnerable to eavesdropping and the like because the input speech waveform itself flows through the communication path. For this reason, there is an attempt to improve the confidentiality of communication by manipulating the voice waveform of analog communication. In Non-Patent Documents 1 and 2, the above attempt is introduced as a prior art. FIG. 8 to FIG. 10 are communication systems in which the confidentiality of analog communication is enhanced by the prior art. In Non-Patent Document 1, in FIG. 8, sampling is performed within a fixed time, and the order of the samples is moved back and forth on the time axis. In FIG. 9, the speech waveform is divided into predetermined blocks and divided into blocks, and confidentiality is obtained by rearranging and transmitting the blocks. In Non-Patent Document 2, as shown in FIG. 10, the spectrum is inverted according to a predetermined rule in the frequency spectrum.
IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. COM-29, NO. I, JANUARY 1981 D. W. Davies (Ed.): Advances in Cryptology-EUROCRYPT '91, LNCS 547, pp. 422-430, 1991.

  However, in these methods, operations such as rearrangement are performed on the sound itself to be encrypted, so the original sound may be partially heard. Then, if you rearrange them using this as a clue, you may be able to restore them relatively easily. As described above, since the search space is limited in the method of performing the operation on the sound itself to be encrypted, the communication cannot be concealed. Therefore, it is a problem that communication is performed with a voice other than the voice to be encrypted to ensure sufficient concealment of communication.

  In order to solve the above problems, first, the present invention provides the following audio system. First, an index string of speech segments is acquired according to the language input input to the system. Next, the index string is replaced with another index string by the encryption key. Speech synthesis is performed using the replaced index sequence to generate a secret speech, and the speech is transmitted through a transmission path. On the receiving side, an index sequence of speech segments is acquired according to the received secret speech. Next, the index string is restored to the index string before transmission with the decryption key. Speech synthesis is performed using the restored index sequence and output as speech.

Specifically, a correspondence table holding unit that holds a correspondence table that associates speech units and indexes, a language input reception unit that receives language input, and a speech element that constitutes a language input from the language input reception unit. A raw index generation unit that generates a raw index that is a single index from a correspondence table, an encrypted index acquisition unit that acquires an encrypted index that is an index obtained by encrypting the raw index, and a voice corresponding to the acquired encrypted index An output unit having a secret speech synthesis unit that acquires a segment from a correspondence table and performs speech synthesis, an output unit that outputs the synthesized secret speech, an acquisition unit that acquires the secret speech, and a voice A correspondence table holding unit that holds a correspondence table in which a segment and an index are associated with each other, and an encryption index that is an index of a speech unit corresponding to the acquired secret speech. An encryption index generator for generating by using a correspondence table box,
A restoration apparatus comprising: a raw index decoding unit that decodes an encrypted index to obtain a raw index; and a language restoration unit that obtains a speech unit from the decoded raw index and restores a corresponding language; An audio system consisting of

  Second, based on the first voice system, a voice system in which the language input to the system is voice is provided.

  Specifically, the language input receiving unit is the voice system according to claim 1, further comprising a voice language input receiving unit that receives an input of a voice language.

  Third, based on the first or second voice system, a voice system in which language input to the system is text data is provided.

    Specifically, the language input receiving unit is the voice system according to claim 1 or 2, further comprising a text language input receiving unit that receives an input of a text language.

  According to the first aspect of the present invention configured as described above, it is possible to acquire an index sequence of speech segments for language input to the system and replace it with an encryption key. Then, by synthesizing speech with the replaced index string, it is possible to flow the secret speech using speech segments other than the input language input to the transmission path. On the receiving side, it is possible to acquire the index of the speech unit according to the received secret speech, restore it with the decryption key, and acquire the index sequence of the original speech unit. If speech synthesis is performed using this index sequence, the original speech can be restored. Thus, the confidentiality of communication is improved by using the phoneme string other than the voice input to the system as the voice that flows through the transmission medium.

  In addition, according to the second aspect of the present invention, since the language input to the system is a voice, it is possible to send the secret voice via the transmission path and to send the voice to the receiving side in almost real time. Thereby, bidirectional real-time communication such as a telephone can be performed while ensuring confidentiality of communication.

Further, according to the third aspect of the present invention, since the language input to the system is text data, it is possible to communicate after converting the text into speech. If a process of converting voice to text data is performed on the receiving side, offline communication such as facsimile can be performed while ensuring secrecy.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. Note that the present invention is not limited to these embodiments, and can be implemented in various modes without departing from the spirit of the present invention.

<Example>
<Overview>
FIG. 1 is a conceptual diagram for explaining an example of the operation of the sound system according to the present embodiment. As shown in this figure, there are a receiver and a transmitter, and the language input of “getting ridiculous” to the transmitter is set as the secret voice “Oyasememe” and transmitted to the receiver. On the receiver side, the secret voice “Oyasememe” is restored, and an output of “becoming candy” is obtained.

  More specifically, after receiving an input of “becoming candy” on the transmitter side, an index of the speech unit database is acquired according to the input pattern. Although an actual speech segment database includes a large number of patterns, for the sake of simplicity, it is assumed that the speech unit database is assigned from an index table as shown in FIG. Then, an index string “A1 G4 E2 E1 I3” can be acquired (FIG. 2B). This index is replaced with another index string based on the encryption generated by the encryption key. Assume that an index string “A5 H1 C4 G1 G4” has been acquired. Then, the secret character string “Oyasememe” is obtained from the index of the segment database (FIG. 2C). The character string is converted into speech by speech synthesis based on the speech segment database and transmitted to the receiver via an analog communication path or the like.

  On the receiver side, an index string is acquired from the speech segment database according to the pattern of the secret speech “Oyasememe”. In this example, an index string “A5 H1 C4 G1 G4” is acquired. When this index string is restored with the decryption key, the index string with “A1 G4 E2 E1 I3” before replacement is acquired at the transmitter. The speech unit database is referred to in accordance with the index string, and the speech corresponding to the character string “Let's become candy” is obtained.

  As described above, the speech system according to the present embodiment has a process of acquiring the index sequence of the speech segment database for the input character string and replacing it with an index sequence corresponding to another speech segment using the encryption key. Is possible. For this reason, even if the secret speech itself is replaced for each segment, it is impossible to obtain the original character string. Therefore, communication is possible while maintaining high confidentiality. Also, since the secret speech is transmitted by voice from the transmitter side to the receiver side using voice synthesis, it is suitable for a simple communication method such as an analog communication path.

<Functional configuration 1>
FIG. 3 is a diagram illustrating an example of functional blocks in the audio system according to the present embodiment. As shown in this figure, the “voice system” of this embodiment includes an “output device” (0301) and a “restoration device” (0302). The output device (0301) includes a language input reception unit (0303), a raw index generation unit (0304), a correspondence table holding unit (0305), an encrypted index acquisition unit (0306), and a secret speech synthesis unit (0307). ) And an output unit (0308). The restoration device (0302) includes an acquisition unit (0309), an encrypted index generation unit (0310), a correspondence table holding unit (0311), a raw index decryption unit (0312), and a language restoration unit (0313). Have

The language input reception unit (0303) has a function of receiving language input in the output device (0301).
A language is data that can be output as speech synthesis such as speech or text. The input can be real-time input with a microphone, input with audio data, input with text data, or the like. The accepted language input is output to the raw index generation unit (0304).

  The raw index generation unit (0304) has a function of generating, from the correspondence table, a raw index that is an index of a speech unit constituting a language input from the language input reception unit. Specifically, the speech unit index information stored in the correspondence table holding unit (0305) is acquired according to the language input. For example, when the language input is speech data, pattern recognition means such as DP matching is used to find the speech unit having the highest suitability for the input speech waveform, and the index information of the segment is obtained. For example, it is acquired from the correspondence table holding unit (0305). The index information obtained here is output to the encrypted index acquisition unit (0306) as an index string.

  The correspondence table holding unit (0305) has a function of holding a correspondence table in which speech units and indexes are associated with each other. Specifically, it is a table that holds a speech unit stored in a speech unit database and a speech unit index in association with each other. For example, if the segment data is composed of 50 sounds, a correspondence table as shown in FIG. As for speech segments, in addition to those composed of only 50 sounds, vowels-vowels, vowels-consonants, consonants-vowels, consonants-consonants, etc. There are various types such as those composed of a plurality of segments and those composed of speech segments at the word level.

  The encrypted index acquisition unit (0306) has a function of acquiring an encrypted index that is an index obtained by encrypting a raw index. The index string as shown in FIG. 2B input from the raw index generation unit (0304) is replaced with another index string (FIG. 2C) with the encryption key. The encryption key is encrypted by using the raw index as input data and replacing it with one of the indexes in the correspondence table holding unit. Various encryption methods such as a public key encryption method and a secret key (common key) encryption method can be adopted. The encrypted index string after the replacement is output to the secret speech synthesizer (0307).

  The secret speech synthesizing unit (0307) has a function of acquiring a speech unit corresponding to the acquired encrypted index from the correspondence table and performing speech synthesis. The encrypted index string after being replaced by the encrypted index acquisition unit (0306) is input, the correspondence table is referred to according to the input encrypted index, and based on this, the speech unit is stored from the speech unit database. To synthesize speech. The synthesized voice is output to the output unit (0308).

  The output unit (0308) has a function of outputting the synthesized secret speech. The output is to output the input voice to the acquisition unit (0309) of the restoration device (0302) via the transmission medium. Although various modes such as analog communication, digital communication, wireless, and wired are conceivable as a transmission method, any method can be adopted as long as it can propagate a voice waveform.

  The acquisition unit (0309) has a function of acquiring the secret speech in the restoration device (0302). The speech waveform data output from the output unit (0308) of the output device (0301) is acquired and output to the encryption index generation unit (0310).

  The encryption index generation unit (0310) has a function of generating an encryption index that is an index of a speech unit corresponding to the acquired secret speech using a correspondence table. Specifically, the speech waveform data from the acquisition unit (0309) is selected by pattern recognition such as DP matching to select the speech unit having the highest suitability, and the corresponding index information is acquired from the correspondence table holding unit (0311). It has the function to do. The index information is output to the raw index decryption unit (0312) as an encrypted index sequence.

  The correspondence table holding unit (0311) has a function of holding a correspondence table in which a speech unit and an index are associated with each other in the restoration device (0302). Reference is made at the time of speech recognition at the encryption index generation unit (0309) or when speech synthesis is performed at the language restoration unit (0313).

  The correspondence table in the correspondence table holding unit (0311) needs to be the same as or a subset of the correspondence table of the correspondence table holding unit (0305) of the output device (0301). The voice system may have a single correspondence table holding unit and share it.

  The raw index decryption unit (0312) has a function of decrypting the encrypted index and acquiring the raw index. Using the encryption method used in the encrypted index acquisition unit (0306) of the output device (0301), it has a function of returning the replaced index string to the original raw index. The obtained raw index is output to the language restoration unit (0313).

The language restoration unit (0313) obtains a speech unit from the decoded raw index,
It has a function to restore the corresponding language. Based on the correspondence table, a speech unit is obtained from the raw index obtained by the raw index decoding unit (0312), and speech synthesis is performed. At this time, the output may be speech or text data corresponding to speech segments.

<Functional configuration 2>
FIG. 4 is a diagram illustrating another example of functional blocks in the audio system according to the present embodiment. As shown in this figure, the “voice system” of this embodiment includes an “output device” (0401) and a “restoration device” (0402). The output device (0401) includes a language input reception unit (0403), a raw index generation unit (0404), a correspondence table holding unit (0405), an encrypted index acquisition unit (0406), and a secret speech synthesis unit (0407). ) And an output unit (0408). The restoration device (0402) includes an acquisition unit (0409), an encrypted index generation unit (0410), a correspondence table holding unit (0411), a raw index decryption unit (0412), and a language restoration unit (0413). Have Since the above-described configuration requirements have already been described above, description thereof will be omitted. A characteristic point of this configuration is that the language input receiving unit (0403) in the “output device” (0401) has “speech language input receiving means” (0414).

  The spoken language input receiving means (0414) has a function of receiving an input of the spoken language. The speech language is not limited to a natural language as long as it can be generated by speech, and may be a singing voice or the like. Which speech language can be used depends on the correspondence table of the correspondence table holding unit (0405) based on the speech segment database.

<Functional configuration 3>
FIG. 5 is a diagram illustrating still another example of functional blocks in the audio system according to the present embodiment. As shown in this figure, the “voice system” of this embodiment includes an “output device” (0501) and a “restoration device” (0502). The output device (0501) includes a language input reception unit (0503), a raw index generation unit (0504), a correspondence table holding unit (0505), an encrypted index acquisition unit (0506), and a secret speech synthesis unit (0507). ) And an output unit (0508). The restoration device (0502) includes an acquisition unit (0509), an encrypted index generation unit (0510), a correspondence table holding unit (0511), a raw index decryption unit (0512), and a language restoration unit (0513). Have Since the above-described configuration requirements have already been described above, description thereof will be omitted. A characteristic point of this configuration is that the language input receiving unit has “text language input receiving means” (0515).

  The text language input receiving means (0515) has a function of receiving text language input. The text language is text data, and any text language can be used as long as speech synthesis is possible. Which text language can be used depends on the correspondence table of the correspondence table holding unit (0505) based on the speech segment database, similarly to the speech language input receiving means (0414).

<Process flow>
FIG. 6 is a flowchart illustrating an example of a process flow in the voice system according to the present embodiment. First, the language input is accepted on the output device side (step S0601). Next, a raw index, which is an index of speech units constituting the language input in the language input accepting step, is generated from a correspondence table in which speech units and indexes are associated (step S0602). Next, an encrypted index that is an index obtained by encrypting the raw index is obtained (step S0603). Next, a speech unit corresponding to the obtained encrypted index is obtained from the correspondence table, and speech synthesis is performed (step S0604). Then, the synthesized secret voice is output (step S0605). On the restoration device side, the secret speech is acquired (step S0606). Next, an encrypted index, which is an index of a speech unit corresponding to the acquired secret speech, is generated using a correspondence table in which the speech unit and the index are associated (step S0607). Next, the encrypted index is decrypted to obtain a raw index (step S0608). Next, a raw index decrypting step for decrypting the encrypted index to obtain a raw index, a speech segment is obtained from the decrypted raw index, and a corresponding language is restored (step S0609).

<Hardware configuration>
FIG. 7 is a schematic diagram showing an example of a configuration in the audio system when the above functional components are realized as hardware. The operation of each hardware component of voice communication will be described using this figure. As shown in this figure, the voice system of the present embodiment includes a transmitter (0701) and a receiver (0702). In the transmitter (0701), a “CPU (central processing unit)” (0703), a “volatile memory” (0704), a “nonvolatile memory” (0705), and a D / A converter that perform various arithmetic processes. (0706), a microphone (0707), and a network I / F (0708). These are connected to each other through a data communication path such as a “system bus” (0709) to transmit / receive information and process information.

In the transmitter (0702), a “CPU (central processing unit)” (0710), a “volatile memory” (0711), a “nonvolatile memory” (0712), and a D / A converter that perform various arithmetic processes. (0713), a speaker (0714), and a network I / F (0715). Then, they are connected to each other by a data communication path such as “system bus” (0716) to transmit / receive information and process information.

  The “volatile memory” (0704, 0711) reads from the “non-volatile memory” (0705, 0712) at the same time that the “CPU” (0703, 0710) executes the program for performing various processes. Provide a work area that is also a work area.

  Here, when there is an audio input to the microphone (0707), data is sent to the CPU (0702) via the D / A converter (0706). In the CPU (0702), A: the raw index generation program held in the volatile memory (0704) is executed, and B: the element DB (0717) stored in the correspondence table and the nonvolatile memory (0705). ), A raw index is created. Next, C: the encryption program is executed by the CPU (0703), and the raw index is replaced with the encryption index of another segment. Using this, the secret speech is generated by executing the D: voice synthesis engine while referring to the segment DB (0717) by the CPU (0703). The voice is transmitted to the receiver (0702) via the network I / F (0708).

  On the receiver side, the secret speech is received at the network I / F (0715). Then, the voice is sent to the CPU (0710), and an encrypted index is generated by the running E: encrypted index creation program. At this time, the program refers to the segment DB (0718) stored in the nonvolatile memory (0712) and the G: correspondence table held in the volatile memory (0711). Next, the encrypted index is returned to the raw index by the F: decryption program. This raw index is voiced by the H: voice synthesis engine while referring to the G: correspondence table and the segment DB (0718). The sound is output from the speaker (0714) via the D / A converter (0713).

<Brief description of effect>
As described above, with the speech system of this embodiment, it is possible to obtain an index sequence of speech segments for the input language in the output device, and replace this index sequence with another index sequence using an encryption key. . This is transmitted through the transmission path. This is received by the restoration device, the original index string is acquired by the decryption key, and the input language can be restored. Thereby, the confidentiality of communication improves.

The figure for demonstrating the outline | summary of the process by the audio | voice system of an Example The figure for demonstrating the outline | summary of the process by the audio | voice system of an Example The figure showing an example of the functional block of the audio | voice system of an Example The figure showing an example of another functional block of the audio system of an example The figure showing an example of another functional block of the audio system of an example The flowchart showing an example of the flow of processing in the audio system of the embodiment Schematic showing an example of a hardware configuration in the audio system of the embodiment Schematic for explaining the encryption method in the prior art Schematic for explaining the encryption method in the prior art Schematic for explaining the encryption method in the prior art

0301 Output device 0302 Restoration device 0304 Raw index generation unit 0306 Encrypted index acquisition unit 0310 Encrypted index generation unit 0312 Raw index decryption unit

Claims (6)

A correspondence table holding unit that holds a correspondence table in which speech units and indexes are associated;
A language input receiving unit for receiving language input;
This is the index of the speech unit that composes the language input from the language input reception unit
A raw index generation unit that generates a raw index from a correspondence table;
An encrypted index acquisition unit that acquires an encrypted index that is an index obtained by encrypting a raw index;
The speech unit corresponding to the obtained encryption index is obtained from the correspondence table,
A secret speech synthesis unit that performs speech synthesis;
An output unit for outputting the synthesized secret voice;
An output device having
An acquisition unit for acquiring the confidential voice;
A correspondence table holding unit that holds a correspondence table in which speech units and indexes are associated;
The encryption index that is the index of the speech segment according to the acquired secret speech
An encrypted index generation unit generated using a correspondence table;
Decrypt encrypted index to get raw index
A raw index decryption unit;
The speech segment is obtained from the decoded raw index,
A language restoration unit that restores the corresponding language;
A restoration device having
An audio system consisting of   The speech system according to claim 1, wherein the language input reception unit includes a speech language input reception unit that receives an input of the speech language.   The speech system according to claim 1, wherein the language input receiving unit includes a text language input receiving unit that receives an input of a text language. A language input reception step for receiving language input;
A raw index generation step of generating a raw index, which is an index of a speech unit constituting a language input in the language input reception step, from a correspondence table in which the speech unit and the index are associated;
An encrypted index acquisition step of acquiring an encrypted index that is an index obtained by encrypting a raw index;
A speech unit corresponding to the obtained encrypted index is obtained from the correspondence table, and a secret speech synthesis step for performing speech synthesis,
An output step for outputting the synthesized secret voice;
An acquisition step of acquiring an anonymized voice;
An encrypted index, which is an index of speech segments according to the acquired secret speech,
An encrypted index generation step that uses a correspondence table that associates speech segments and indexes;
A raw index decryption step of decrypting the encrypted index to obtain a raw index;
A language restoration step of obtaining a speech unit from the decrypted raw index and restoring a corresponding language;
A method of operating a voice system comprising:   6. The operation method of an audio system according to claim 5, wherein the language input receiving step includes an audio language input receiving substep for receiving an input of the audio language.   The speech system operating method according to claim 5 or 6, wherein the language input receiving step includes a text language input receiving sub-step for receiving an input of a text language.

Images