JPH09214636A - Method and device for voice communication imbedded with data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To facilitate ciphering of data such as character information and error correction by setting data of coded character information or the like to low-order bits of a data imbedded voice signal so as to attain communication of digital voice information imbedded with the data. SOLUTION: A sampling means 2 samples an output 101 of a low pass filter 1 by using a frequency signal of 8kHz. A quantization means 3 quantizes a sampled value 102 in 128 steps to generate a quantization output 103. A character information input means 4 is an input means for character codes coded in 8-bit. A coding means 5 applies 8-bit binary coding to the output of the quantization means 103 to generate binary coded voice data. Low-order n-bits (n is an integer being one or over and 7 and lower, practically 4 or below) of the 8-bit voice code are replaced with n-bits in 8-bit character code 104. The character information imbedded voice code is generated through the replacement and outputted as a character imbedded voice signal 105.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data-embedded voice communication method and device for embedding data in which characters and the like are encoded in digitally encoded voice for communication.
In particular, the present invention relates to a data embedded voice communication method and apparatus applying a so-called dither technique introduced to decorrelate the unavoidable quantization noise with the analog voice input when digitizing the input analog voice. .

[0002]

2. Description of the Related Art As a multimedia environment is prepared, a system is being provided in which information such as images, sounds, and texts can be easily loaded into a computer and transmitted / received.
Conventionally, in multimedia processing, each media is treated as an independent file, but if these can be fused or integrated, the handling becomes simple. From such a point of view, it has already been attempted to embed character information in image information having a high degree of redundancy. For example, the present inventor, K. Matsui, Morikita Publishing, 1993.
In the year, detailed description is given in a technical document called "image deep encryption" [hereinafter referred to as Document 1].

Although there are few literatures regarding the technology of embedding character information in voice information, for example, R. Steele, D. Vitello, "Simul.
taneous Transmission of Speech and Data Using Code
-Breaking Techniques ", BSTJ, Vol.60, No.9, pp.2081-
2105 (1981) [hereinafter referred to as reference 2] and WC Wong, R. St.
eele, CSXydeas, "Transmitting Data on the Phaseof
Speech Signals ", BSTJ, Vol.61, No.10, pp.2947-2970
(1982) [hereinafter referred to as Reference 3]. These Documents 2 and 3 propose a method of simultaneously transmitting character information in voice communication using a telephone, but these are one example of synthesizing voice information and character information in an analog system.

[0004]

In the synthesis of the voice information and the character information shown in Documents 2 and 3, the synthesis is performed when both of the information are in an analog state, and therefore the encryption for confidentiality of information in communication is performed. It is difficult to convert the information, and error correction is difficult if the character information changes due to noise or the like in the communication path. Encryption of character information and error correction are facilitated by digitizing communication information. Therefore, it is an object of the present invention to provide a method and apparatus for embedding digital data in digital voice information for communication.

[0005]

In order to solve the above-mentioned problems, the present invention provides the following means.

A sample value of the analog voice signal is generated by sampling the analog voice signal, the sample value is converted into a discrete value by quantizing the sample value, and the discrete value is converted into m ( m is an integer of 2 or more) converted into a code of
An audio signal in which the data is embedded is generated by replacing the lower n bits (n is an integer of 1 or more and less than m) of the bit code with bits of data having no correlation with the audio signal, and the data is embedded. A voice communication method for transmitting a voice signal from a transmitting device to a receiving device via a communication path,
The lower n bits in the data-embedded voice signal are data obtained by encoding character information, image information, voice information other than the voice signal, a computer program list, or the like, or bits forming a part of the data. A data-embedded voice communication method characterized by:

In a data-embedded voice communication method of embedding data in voice information to generate data-embedded voice information and communicating the data-embedded voice information, the voice information is coded with m (m is an integer of 2 or more) bits. , The lower n bits of the m-bit speech code (n is 1
By replacing the above (an integer less than m) with a code representing the data, a voice code in which the data is embedded is generated,
A data-embedded voice communication method for transmitting the data-embedded voice code from a transmitter to a receiver via a communication path.

Sampling means for generating a sample value of the analog audio signal by sampling the analog audio signal; and quantizing means for converting the sample value into a discrete value by quantizing the sample value. The discrete value is m (m is an integer of 2 or more)
Converted into a bit code, and the lower n of the m-bit code
An encoding means for generating an audio signal in which the data is embedded by replacing a bit (n is an integer of 1 or more and less than m) with a bit of data having no correlation with the audio signal; Modulation means for generating a modulated wave by modulating a carrier wave and sending the modulated wave to a communication path, and data obtained by encoding character information, image information, voice information other than the voice signal, computer program list, etc. And a decoding means for receiving the modulated wave from the communication path and decoding the data-embedded voice signal from the modulated wave, and a means for inputting the data-embedded voice signal into an analog voice signal. Digital to convert
Analog receiving means is provided on the receiving side, and the encoding means sets the bits forming the data to the lower n bits,
The data embedded voice communication device, wherein the decoding means decodes the data from the lower n bits of information in the decoded data embedded voice signal.

[0009] A transmitter for embedding data in voice information to generate data-embedded voice information, and transmitting the data-embedded voice information to a communication channel, and a device for receiving the data-embedded voice information via the communication channel, In a data-embedded voice communication device including a receiving device that outputs information and data separately, the transmitting device represents the voice information by a code of m (m is an integer of 2 or more) bits, and the m-bit Lower n bits of speech code (n is 1 or more and m
(Integer less than) is replaced with a code representing the data to generate a voice code in which the data is embedded, and the data-embedded voice code is transmitted to the communication path, and the receiving device transmits the data-embedded voice code to the communication. From the channel, reproduces the data with the lower n bits of the data-embedded voice code, outputs the data, converts the data-embedded voice code into an analog voice signal, and uses the analog voice signal as the voice information. A data embedded voice communication device characterized by outputting.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention is a method and apparatus for digitizing an analog voice signal to encode voice, and originally encoding to embed data handled by a computer in digital voice information. In this specification, characters,
Information obtained by encoding images, sounds, computer program lists, etc. is called data. The practical code is
It is a binary code. The human voice is used as the voice information.

First, let's consider borrowing a part of the audio digital signal to embed the data and store or transmit the data. By this embedding process, only the audio signal is transmitted on the surface, and the data is dispersed and concealed, so that the concealment effect is brought about. It should be noted that the audio signal acting as a carrier of embedded information is also data because it is digitized, but in this specification, it is transmitted by being embedded in the digital audio signal instead of the lower bit of the digital audio signal. Only the information that is available is called data. This embedding process is a process of adding noise to the audio signal if the audio signal is regarded as the main signal, similar to the method of dithering the audio signal at the time of quantization.

FIG. 1 is a block diagram conceptually showing a character information embedded voice communication device which is an embodiment of the "data embedded voice communication device" of the present invention. The apparatus shown in FIG. 1 is realized by a personal computer or a workstation.

An analog wave uttered from a human oral cavity is input from a microphone into a computer and processed according to FIG. Since voice waves are converted into digital signals,
Compared to analog signals, digital processing can be performed easily. Assuming that this input frequency is 300 to 3400 Hz, the analog input 100 is quantized by 8 bits.

In the device of FIG. 1, the analog input 100 sent from the microphone is a low-pass filter 1 of 3400.
The components above Hz are removed. The sampling means 2 samples the output 101 of the low-pass filter 1 at 8 kHz. The quantizing means 3 quantizes the sample value 102 into 128 steps and generates a quantized output (discrete value) 103. The character information input means 4 is an input means of a character code encoded by 8 bits. In the encoding means 5, first, the quantized output 103 is subjected to 8-bit (m = 8) binary encoding to generate binary encoded audio data. Next, the lower n bits (n is an integer of 1 or more and 7 or less. In practice, n is 4 or less) of the 8-bit speech code are replaced with n bits of the 8 bits forming the character code 104. . By this replacement, the character information embedded voice code is generated and output as the character information embedded voice signal 105. Frame construction means 6
Then, a frame sync pulse is added to the 8-bit character information embedded voice signal 105 to generate a framed digital output 106, and the digital output 106 is sent to the communication path. In the channel disassembling means 7, the digital output 106 multiplexed in the communication path
Is received from the communication channel, the synchronization pulse of the channel assigned to itself is identified, and the digital input 1
07 is extracted. The decoding means 8 decodes the character information embedded voice code from the digital input 107 and outputs it as the character information embedded voice signal 108. In the decoding means 8, the character code 110 is constructed from the lower n bits of the character information embedded voice code. The character information recording means 10 stores the character code 110 and prints it on paper or displays it on a display. The low pass filter 9 converts the character information embedded voice signal 108 into an analog voice signal and outputs it as an analog output 109.

The outline of the operation of the character information embedded voice communication device of FIG. 1 has been described above. The operation principle of the device of FIG. 1 will be described in detail below.

Usually, due to the quantization, the error between the input analog signal and the quantized output becomes the quantized distortion, which becomes a noise generation source. This error component becomes white noise in a signal with a large change in the input signal, but for a signal with a low input level and a small number of quantization steps, or a signal that changes very slowly, the quantization noise has a strong correlation with the input. There will be some distortion. For example, the relationship is shown in FIG.

When the sine wave of FIG. 2 (a) is used as an input signal and is sampled at a sampling frequency twice that of the input signal and quantized, the quantized output becomes a square wave as shown in FIG. 2 (b).
The waveform of the difference between the waveform of (b) and the waveform of (a) becomes the quantization noise of FIG. This quantization noise has a strong correlation with the input sine wave. The sine wave in FIG. 2A is an example of a signal having a small number of quantization steps. In this example, because of the relationship with the sampling frequency, the sinusoidal input signal is as if it were a square wave input, and the quantized output has pulsed quantization noise and contains strong level harmonics. I'm out.

FIG. 3A shows the relationship between the input level (horizontal axis) to the quantizing means and the quantized output (vertical axis).
(B) shows the relationship between the input level (horizontal axis) and the quantization noise level (vertical axis). FIG. 3 shows that when the input signal level changes gradually, strong quantization noise is generated for each quantization step width Δ.

Regarding reduction of quantization noise by dither, Ohga, Yamazaki, Kaneda: Acoustic Systems and Digital Processing, The Institute of Electronics, Information and Communication Engineers, Chapter 3, pp.55-71 (1995) [hereinafter, referred to as reference 4
4], and FIG. 6 and FIG.
Is a diagram published in Reference 4. FIG. 4 is a waveform diagram displayed on an oscilloscope with the horizontal axis representing time and the vertical axis representing level. FIG. 4A shows the same sinusoidal signal as in FIG. 2A, which is the input signal to be quantized. FIG. 6B shows a dither signal. The dither signal is white noise whose amplitude is about 1/2 to 2 times the quantization step width Δ of the input signal. FIG. 4C shows a signal obtained by adding the dither signal shown in FIG. 4B to the input signal shown in FIG. FIG. 4 (d)
Is a waveform observed when a signal obtained by quantizing the analog signal of FIG. 7C is displayed on an oscilloscope. FIG.
(E) is a signal obtained by analog-converting the signal shown in (d) of FIG. FIG. 4 (f) is a diagram showing the quantization noise included in the signal of FIG. 4 (e).

FIG. 5A shows the input signal as shown in FIG. 4C by adding the dither signal shown in FIG. 4B to the input signal.
It is a figure which shows the relationship between an input level (horizontal axis | shaft) and a quantized output (vertical axis) when it quantizes after dithering like this. Further, FIG. 5B is a diagram showing the relationship between the input level (horizontal axis) and the quantization noise (vertical axis) in the quantized output.

FIG. 6 (a) shows the relationship between the input level (horizontal axis) and the dither subtraction signal (vertical axis) corresponding to FIG. 4 (e). FIG. 4B shows the input level (horizontal axis) and FIG.
The relationship with the quantization noise corresponding to (f) is shown.

As shown in FIGS. 4 (f) and 6 (b), the quantization noise does not correlate with the input signal due to the dithering, and the quantized output has no harmonic component of the input signal. rare.

As described above, the dither signals having the amplitude of about 1/2 to 2 times the quantization step width Δ of the input signal are added,
By randomizing the quantized output, the input signal and the quantization noise can be uncorrelated. By this processing, the input value and the expected value of the quantized output can be matched. The purpose of this dither introduction is not to mask the noise, but to de-correlate the input and quantization noise,
That is, the suppression of harmonics in the quantized output.

In the present invention, data such as character information is used in place of the dither signal to suppress the harmonic components of the input signal from appearing strongly in the quantization noise. Since the amplitude of the dither signal is about 1/2 to 2 times the quantization step width Δ,
It is preferable that the ratio of the data to the voice information is similar.

In the embodiment of FIG. 1, the quantized output 103 which is the output of the quantizing means 3 is 8-bit binary encoded by the encoding means 5. That is, an 8-bit binary voice code is generated. The character information is input from the character information input means 4.
The 8-bit binary coded data by the IS (Kanji) code, that is, the character code 104 is given to the coding means 5.
The character code 104 represents one character with 8 bits, and is input to the encoding means 5 as a code sequence in which 8-bit codes are continuous. The encoding means 5 replaces the lower n bits of the 8-bit voice code with n bits sequentially extracted from the code sequence of the character information. This replacement is performed every n bits from the beginning to the end of the code sequence of the character information to be transmitted. Replacing the lower n bits of the voice code with n bits of the character code sequence is called embedding the character code in the voice code or embedding the character information in the voice information.

[0026]

EXAMPLE A workstation (S
It was carried out in UNSPARC station 10). Four types of samples as voice input (male music, female music,
(Male announcement, white noise) was prepared, and character information of 1 to 7 bits was embedded in each sample point (n = 1 to 7). FIG. 7 shows a comparison of the input / output S / N ratio with respect to the number of embedded bits for each sample.

From this result, it can be seen that the S / N ratio deteriorates almost linearly with respect to the embedded bit number n regardless of the type of sample. In terms of reproduced sound quality, it was confirmed that noise was not adversely affected as noise by the caller of the telephone if it was within 3 bits / sample point (n ≦ 3).

In FIG. 8, n = 3 in this embodiment,
6 is a diagram showing a display by the signal processor when the level of the analog input 100 is plotted on the horizontal axis and the character information embedded voice signal 108 is plotted on the vertical axis, and is a diagram showing a state of dithering in this embodiment. FIG. As can be seen from the figure, in this embodiment, the lower 3 bits of the 8-bit voice code are replaced with 3 bits sequentially extracted from the character code string having no correlation with the voice code. Therefore, the discrete value (vertical axis) fluctuates (steps) at various positions of the audio input signal level (horizontal axis).

In FIG. 9, in the present embodiment, n = 3, time is plotted on the horizontal axis, and the quantized output 103, the character information embedded voice signal 108 (displayed as “embedded output” in FIG. 9) and the character information embedded voice signal. 10 is a characteristic diagram in which the lower 3 bits (character code component, shown as “noise component” in FIG. 9) in 108 are plotted along the vertical axis to indicate the degree of noise given to the voice information by the embedded character information. In this figure, the vertical axis represents the standardization level, and the horizontal axis has a length of 1/80.
The time is 00 seconds. From FIG. 9, even if the lower 3 bits of the 8-bit voice code are replaced with the 3 bits of the character code string,
It can be seen that the voice code is not significantly affected and the noise level is sufficiently lower than the voice level for practical use.

FIG. 10 is a diagram showing the distribution of frequency components when n = 4, that is, when a 4-bit character code string is embedded in an 8-bit voice code in this embodiment. This frequency component distribution is the quantized output 103
And analog output 109 (“embedding result” in FIG. 10)
It is generated by FFT (Fast Fourier Transform). From this figure, it can be seen that there is no change in the frequency component before and after the embedding of the character code.

From the description so far made with reference to FIGS. 7 to 10, even if the present invention is applied to secretly embed data such as character information using a voice signal as a carrier, if the number of bits to be embedded is small, the sound quality, It was confirmed that no inconvenience was felt in terms of frequency spectrum and hearing.

In the embodiment described above, the binary code of the JIS code is converted into the character code 104 from the character information input means 4.
Is given to the encoding means 5, and the encoding means 5 sequentially extracts n bits from the character code string consisting of the continuous character code 104, and replaces it with the lower n bits of the m-bit speech code, Information is embedded in voice information.
However, embedding a known code such as a JIS code in the voice code as it is in the encoding means 5 is not preferable from the viewpoint of ensuring the confidentiality of communication. Therefore, the present invention can be applied to a case where the encoding unit 5 encrypts a character code made of a binary code by a random number key system to improve the confidentiality of communication.

Further, in order to prevent the character code from disappearing or an error due to noise mixed in the communication path, the character code is error-corrected before being embedded in the voice code in the encoding means 5. It is also possible in the present invention to embed after converting into a code sequence.

As described above, in the present invention, the character code to be embedded in the voice code includes an encrypted code and an error correction code sequence. The present invention can also be applied to communications that employ a data compression technique that takes into consideration high efficiency encoding and the like. Further, in the above-mentioned embodiments and examples, the data to be embedded is character information, but in the present invention, not only the character information but also image information such as FAX data, voice information other than voice signals as a carrier, computer Needless to say, communication can be performed by embedding data obtained by encoding a program list or the like.

[0035]

As described above in detail with reference to the embodiments and examples, the present invention embeds data such as digital character information in digital voice information and communicates data in voice communication such as telephone. It is possible to do together. In the present invention, since data such as character information is embedded in a digital format for communication, encryption of data such as character information and error correction are easy.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an embodiment of the present invention.

FIG. 2 is a diagram showing an example in which, in quantization, an input harmonic component is included in quantization noise at a strong level.

FIG. 3 is a diagram showing a relationship between an input level (horizontal axis), a quantized output, and a quantized noise (vertical axis) in the quantizing means.

FIG. 4 is a diagram illustrating an example of dithering a sine wave with a dither signal.

5 is a diagram showing the quantized output of FIG. 4D and the quantization noise (vertical axis) in the quantized output in relation to the signal level (horizontal axis).

FIG. 6 is a diagram showing the dither subtraction output of FIG. 4E and the quantization noise (vertical axis) in the dither subtraction output in relation to the signal level (horizontal axis).

FIG. 7 is a diagram illustrating embedded bit numbers n and S in the embodiment of FIG.
The figure which shows the relationship of the deterioration degree of / N.

8 is a diagram illustrating the embedded bit number n of 3 in the embodiment of FIG.
FIG. 7 is a diagram showing the relationship between the quantized output 103 (input signal, horizontal axis) and the character-information-embedded audio signal 108 (code output, vertical axis).

9 is a diagram illustrating the embedded bit number n of 3 in the embodiment of FIG.
, The quantization output 103, the character information embedded voice signal 108 and the character code 110 (“noise component”)
Is displayed) and the vertical axis represents time, and the horizontal axis represents time.

10 is a diagram showing the distribution of frequency components in the character information embedded voice signal 108 when the embedded bit number n is 4 in the embodiment of FIG.

[Explanation of symbols]

1 ... Low-pass filter 2 ... Sampling means 3 ... Quantization means 4 ... Character information input means 5 ... Encoding means 6 ... ..Frame construction means 7 ... Channel decomposition means 8 ... Decoding means 9 ... Low-pass filter 10 ... Character information recording means 100 ... Analog input 101・ ・ ・ ・ ・ Low-pass filter output 102 ・ ・ ・ ・ ・ Sample value 103 ・ ・ ・ Quantization output 104 ・ ・ ・ ・ ・ Character code 105 ・ ・ ・ ・ ・ Character information embedded voice signal 106 ・ ・ ・・ Frame configuration Character information embedded voice signal (digital output) 107 ・ ・ ・ Digital input 108 ・ ・ ・ Character information embedded voice signal 109 ・ ・ ・ Analog output 110 ・ ・ ・ Character code

Claims (4)

[Claims] 1. A sample value of the analog audio signal is generated by sampling the analog audio signal, the sample value is converted into a discrete value by quantization of the sample value, and the discrete value is converted into a discrete value. m
(M is an integer greater than or equal to 2) bit code, and the lower n bits (n is an integer greater than or equal to 1 and less than m) of the m-bit code are data bits that have no correlation with the audio signal. A voice communication method of generating a voice signal in which the data is embedded by replacement and transmitting the data-embedded voice signal from a transmitter to a receiver via a communication path, wherein the lower n bits of the data-embedded voice signal are A data-embedded voice communication method comprising: character information, image information, voice information other than the voice signal, data obtained by encoding a computer program list, or bits constituting a part of the data. 2. A data-embedded voice communication method for generating data-embedded voice information by separately embedding data in voice information, and communicating the data-embedded voice information, wherein the voice information is m (m is an integer of 2 or more). ) Bit code, and the lower n bits of the m-bit speech code (n is
A data-embedded voice that generates a voice code in which the data is embedded by replacing (an integer of 1 or more and less than m) with a code representing the data, and transmits the data-embedded voice code from the transmission device to the reception device via the communication path. Communication method. 3. Sampling means for generating a sample value of the analog audio signal by sampling the analog audio signal, and quantizing means for converting the sample value into a discrete value by quantizing the sample value. And the discrete value is converted into a code of m (m is an integer of 2 or more) bits, and the lower n bits (n is an integer of 1 or more and less than m) of the m-bit code are converted into the audio signal. And encoding means for generating an audio signal in which the data is embedded by replacing it with a bit of data having no correlation,
Modulating means for generating a modulated wave by modulating a carrier wave with the data-embedded audio signal and sending the modulated wave to a communication path;
Text information, image information, information of voice other than the voice signal,
A means for inputting data obtained by encoding a computer program list or the like on the transmission side, a decoding means for receiving the modulated wave from the communication path and decoding the data-embedded voice signal from the modulated wave; The receiving side is provided with a digital / analog converting means for converting the data-embedded voice signal into an analog voice signal, the encoding means sets the bits forming the data to the lower n bits, and the decoding means decodes the decoded data. A data-embedded voice communication device, wherein the data is decoded from the lower n bits of information in the data-embedded voice signal. 4. A transmission device for embedding data in voice information to generate data-embedded voice information, and transmitting the data-embedded voice information to a communication channel, and receiving the data-embedded voice information via the communication channel, In a data-embedded voice communication device comprising a reception device that outputs the voice information and data separately, the transmission device represents the voice information by a code of m (m is an integer of 2 or more) bits, and m A voice code in which data is embedded is generated by replacing the lower n bits (n is an integer of 1 or more and less than m) of the bit voice code with a code representing the data, and the data-embedded voice code is transmitted to the communication path. And the receiving device receives the data-embedded voice code from the communication channel and reproduces the data by the lower n bits of the data-embedded voice code. , And outputs the data, the data embedding audio code into an analog audio signal, the data embedding voice communication device and outputs the analog audio signal as the audio information.