JPH0433176B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a multiplex communication device, and more particularly to a multiplex communication device that transmits a voice signal by multiplexing a predetermined symbol signal with the voice signal.

Generally, communication devices that transmit voice signals include telephones, which are closely used in daily life.
It is used as an extremely important means of information transmission in a wide range of applications through wired communication, wireless communication, optical communication, etc. These various communication devices are strongly required to expand their respective transmission capacities in response to the increased demand for information transmission and their increased use. This request for expanding information transmission capacity has a wide variety of countermeasures, depending on the operating area and scale of the information transmission system.

The multiplex communication device of the present invention also focuses on the nature of the predictive characteristics of a voice signal, and in a communication device whose purpose is to transmit a voice signal, a predetermined symbol signal is multiplexed onto the voice signal through a special method. The purpose of this is to contribute to expanding the information capacity by transmitting information.

A multiplex communication device of the present invention is a communication device that transmits a predetermined audio signal, and in order to multiplex and transmit a predetermined symbol signal to a transmission line of the communication device, the multiplex communication device converts the audio signal from analog to digital. an A-D converter that outputs a digital audio signal, inputs the digital audio signal, and converts the digital audio signal into a multiplexed digital audio signal with some sampled values missing through a buffer memory by the control action of the symbol signal; A modulation processor converts and outputs the multiplexed digital audio signal outputted from the modulation processor.
A speech prediction device comprising: a D-A converter that performs A-conversion and outputs a multiplexed analog audio signal; and a low-pass filter that removes high-order frequency components contained in the multiplexed analog audio signal. an A-D converter comprising a characteristic conversion means on a transmission side, which receives a multiplexed analog audio signal generated by the voice prediction characteristic conversion means, performs AD conversion, and outputs a multiplexed digital audio signal; a window processor that inputs a multiplexed digital audio signal and cuts out and outputs the multiplexed digital audio signal corresponding to an analysis section of a predetermined time length arranged at a time difference of a predetermined sampling period via a buffer memory; ,
A linear method that inputs the multiplexed digital audio signal output from the window processor, performs linear prediction analysis for each analysis interval, and extracts and outputs the linear prediction coefficient of the multiplexed digital audio signal in each analysis interval. A prediction analyzer inputs the linear prediction coefficients, calculates and outputs the spectral distance between the mutually adjacent analysis sections, and also eliminates discontinuities caused by dropping of sampled values in the modulation processor. The waveform is corrected by inputting the distance measuring device that detects and outputs it, the multiplexed digital audio signal outputted from the window processor, and the spectral distance information, and referring to the time position corresponding to the point where the sampled value drops. multiplexed digital audio signal,
A demodulation processing extraction means is provided on the receiving side, and includes a waveform correction processor that sends data to the linear prediction analyzer as a target for waveform prediction analysis.

Hereinafter, the present invention will be described in detail with reference to the drawings. FIGS. 1a and 1b show an example of a voice prediction characteristic converting means on the transmitting side and a demodulation processing extracting means on the receiving side, respectively, which are essential requirements of the present invention. FIG. 2 is a block diagram showing the main parts.

As shown in FIG. 1a, the transmitting side speech prediction characteristic converting means 5 of the present invention includes an A-D converter 1,
The receiving side demodulation processing extraction means 11 of the present invention includes a modulation processor 2, a DA converter 3, and a low-pass filter 4, and as shown in FIG. It includes an AD converter 6, a window processor 7, a waveform correction processor 8, a linear prediction analyzer 9, and a spectral distance measuring device 10.

In FIG. 1a, an audio signal input from a terminal 101 is sampled and A-D converted using a sampling frequency of 8 KHz in an A-D converter 1, and output as a series of digital audio signals. and is input to the modulation processor 2. In the modulation processor 2, the series of digital audio signals are stored in a predetermined buffer memory. A predetermined symbol signal to be multiplexed is simultaneously input to the modulation processor 2 from a terminal 102, and the digital audio signal is read out from the buffer memory under the control of the input of this symbol signal. Certain pulses among the 8KHz readout pulses are removed through a predetermined procedure. Therefore, the digital audio signal read from the buffer memory and output from the modulation processor 2 is
The sampled value corresponding to the specific read pulse is output in a state where it is omitted. This digital audio signal is subjected to D-A conversion in a D-A converter 3,
Unnecessary high-order frequency components are removed via a low-pass filter 4, and the signal is output via a terminal 103 as a multiplexed audio signal based on the symbol signal.

What is shown in FIGS. 3a and 3b are the correspondence between the audio signal input from the terminal 101 and its sampling extraction point, and the outputs of the D-A converter 3 and the low-pass filter 4, respectively. 2 is a diagram showing the correspondence between the multiplexed audio signal and sampling points when the multiplexed audio signal is read out from the converter 2. FIG.

In FIG. 3a, the input audio signal 501 is
Although sampled at the sampling period T (corresponding to the aforementioned 8KHz), when read out from the modulation processor 2, as shown in FIG. 3b, in this example, the symbol signal " In the time interval corresponding to 1'', the sampled value corresponding to the read pulse 503 is omitted.

Furthermore, in this example, the sampled values are not deleted in the time interval corresponding to the symbol "0" and are read out normally. That is, in this embodiment, the audio signal 501 is multiplexed by the audio prediction characteristic conversion means 5 in such a manner that the sampled value corresponding to a specific readout pulse is deleted under the control of the symbol signal as described above. is converted into an audio signal 502. This conversion effect means conversion of the predicted characteristics for the audio signal.
This multiplexed audio signal is transmitted to the receiving side via various transmission paths such as wired, wireless, and optical lines.

Next, on the receiving side, in FIG. 1b,
On the one hand, the multiplexed audio signal input from terminal 104 is sent to terminal 1 as an original audio signal.
05, and on the other hand, it is input to an A-D converter 6 where it is A-D converted and input to a window processor 7 as a digitized multiplexed audio signal. In the window treatment device 7,
This digitized multiplexed audio signal is temporarily stored in a buffer memory, and is then sequentially set at time differences corresponding to a predetermined sampling period, and waveforms corresponding to analysis sections having a predetermined time length are cut out. read out via. On the one hand, this read multiplexed audio signal is input to the linear prediction analyzer 9, where it is subjected to linear prediction analysis for each of the analysis sections. In general, speech waveforms can be predicted linearly, and the following simultaneous linear equations hold true for the prediction coefficient α _i (i=1, 2, . . . , p) in each analysis interval.

φ ₁₁ φ ₂₁ φ ₃₁ …φ _p1 φ ₁₂ φ ₂₂ φ ₃₂ …φ _p2 φ ₁₃ φ ₂₃ φ ₃₃ …φ _p3 〓 〓 〓 〓 〓 〓 φ _1p φ _2p φ _3p ……φ _pp α ₁ α ₂ α ₃ 〓 〓 α _p =φ ₀₁ φ ₀₂ φ ₀₃ 〓 〓 φ _0p (1) In the above equation, φ _ij (i=1, 2,..., p; j=
1, 2,..., p) and φ _0j (j=1, 2,...,
p) are covariance coefficients given by the following equations, where S(x) is the audio waveform.

That is, the linear prediction coefficient α _i for the speech waveform S(x) is determined from the above equation (1).

Through the linear prediction method as described above, the linear prediction analyzer 9 performs waveform analysis on the multiplexed audio signal for each of the above-mentioned analysis intervals. The correspondence relationship between the multiplexed audio signal and each analysis section in this case is as shown in FIG.

In FIG. 4, for the multiplexed audio signal 504, analysis sections [1], [2], [3], [4], . . .
[k], [k+1], [k+2], ..., [k+r], [
k
+r+1], ... correspond, and the time difference between adjacent analysis intervals is equal to the sampling period T. Furthermore, in response to the predictive characteristic conversion effect on the audio signal by the aforementioned audio predictive characteristic converting means 5, the sampled value is removed at a point 506 on the waveform of the multiplexed audio signal 504. Therefore, the linear predictive analyzer 9
When waveform prediction analysis is performed for each analysis interval in , as is clear from Fig. 4, the sampled value is always present in each analysis interval [k+1], [k+2], ..., [k+r]. The target is the waveform time domain that has not been removed.

Now, let the audio waveform to be linearly predicted be S(x), and the sampled sequences of this audio are S(-m), S(-m+1),...
..., S(-p), S(-p+1),..., S(-l+
1), S(-l), S(-l+1), ..., S(-1)
,
Let S(0), S(1), S(2), ... and S(-
If the covariance coefficient is calculated for the series with l) removed, the following equation is obtained.

φ ₀₁ ′=φ ₀₁ +S(-l+1)・S(-l-1) +S(-n-1)・S(-n-2) −S(-l+1)・S(-l) −S(- l)・S(−l−1) (3) That is, φ ₀₁ ′≠φ ₀₁ (4) Similarly, the following equation generally holds true.

φ _0j ′≠φ _0j φ _ij ′≠φ _ij (5) Therefore, the linear prediction coefficient α′ _i (i=1, 2, ..., p) obtained by the following equation (6) is Linear prediction coefficient α _i (i=1, 2, ..., p) obtained by equation (1)
and have different values.

φ′ ₁₁ φ′ ₂₁ φ′ ₃₁ ...φ′ _p1 φ′ ₁₂ φ′ ₂₂ φ′ ₃₂ ...φ _{′ p2} φ′ 13 φ′ ₂₃ φ′ ₃₃ ...φ′ _p3 φ′ _1p φ′ _2p φ′ _3p ……φ′ _pp α′ ₁ α′ ₂ α′ ₃ 〓 〓 α′ _p =φ′ ₀₁ φ′ ₀₂ φ′ ₀₃ 〓 〓 φ′ _0p (6) In general, the linear prediction coefficient α _i is the covariance coefficient It is much more sensitive to discontinuities in the audio waveform than φ _0j and φ _ij themselves, and the above α′ _i has a large deviation from the above α _i .

In the linear prediction analyzer 9, the spectral distance measuring device 10 calculates the spectral distance D _k between adjacent analysis intervals, corresponding to the linear prediction coefficients obtained for each analysis interval. The spectral distance D _k is given by the following equation.

D _k = _p 〓 ⁱ⁼¹ W _i (α _i ^(k) − α _i ^(k+1) ) ² (7) In the above equation, W _i is the spectral sensitivity.
Analysis sections [1], [2], [3], [4], ..., [k], [corresponding to the speech waveform 504 shown in FIG. 4]
k+
1], [k+2], ..., [k+r], [k+r+1]
,
An example of the spectral distance D _k calculated between adjacent analysis sections for ... is shown in FIG. As is clear from FIG. 5, the difference between adjacent analysis sections corresponding to analysis sections [k+1], [k+2], ..., [k+r] whose analysis target area is the sampling value dropping point 506 in FIG. It can be seen that the spectral distance takes a significantly larger value than in the case of other analysis intervals.

In the spectral distance measuring device 10, the discontinuity that occurs due to the dropout of the sampled value is detected by measuring the spectral distance, and is transmitted to the waveform correction processor 8 together with the spectral distance measurement information. In the waveform correction processor 8, the multiplexed audio signal outputted from the window processor 7 and the spectral distance measurement information are input, and a sample corresponding to the sampling value dropout point extracted from the spectral distance measurement information is input. With reference to the time position of the multiplexed audio pulse, waveform correction processing is performed on the multiplexed audio signal, and the resultant signal is output to the linear prediction analyzer 9. In the linear prediction analyzer 9,
A linear prediction analysis is performed on the waveform-corrected multiplexed audio signal input from the waveform correction processor 8 using a method similar to that described above, and linear prediction coefficients are extracted and sent to the spectral distance measuring device 10.

In the spectral distance measuring device 10, the spectral distance between adjacent analysis sections is extracted from the linear prediction coefficient, as in the case described above. The spectrum distance obtained in this case is because the analysis target in the linear prediction analyzer 9 is a multiplexed audio signal whose waveform has been corrected in the waveform correction processor 8.
As shown in FIG. 5, analysis interval [k+1],
As seen in [k+2], ..., [k+r],
The spectral distance never takes an abnormally large value. As a result, the spectral distance measuring device 10
In this method, it is possible to accurately verify the time position of a discontinuous point on the waveform of a multiplexed audio signal extracted from spectral distance information calculated by waveform analysis of the multiplexed audio signal before correction. A specific identification signal is generated corresponding to the time position of the discontinuous point and outputted via the terminal 106. That is,
The symbol time interval from which the sampled value controlled by a predetermined symbol signal has been removed by the voice prediction characteristic converting means 5 on the transmitting side is clearly identified and divided by the demodulation processing extracting means 11 on the receiving side.
For example, in the case of the multiplexed audio signal shown in FIG. has not been carried out.
In this case, it is clear that the demodulation processing extraction means 11 outputs a specific identification signal in response to the symbol "1" and does not output anything in response to the symbol "0". That is, by multiplexing the symbol signal with respect to the original audio signal through the audio prediction characteristic conversion function, the symbol signal can be superimposed and transmitted in addition to the original audio signal.

Next, another embodiment shown in FIG. 2 will be described. In FIG. 2, the present invention includes a voice prediction characteristic converting means 5, a transmitter 12 on the transmitting side,
The receiving side includes a receiving antenna 14, a receiver 15, and a demodulation processing extraction means 11. On the transmitting side, a voice signal and a predetermined symbol signal are inputted from terminals 107 and 108, respectively, and a predetermined multiplexed voice is inputted from the voice prediction characteristic converting means 5 through the voice prediction characteristic conversion function described above. The signal is output and input to the transmitter 12 as a baseband signal.

The transmitter 12 generates a transmission signal using a predetermined modulation method, and sends it to the receiving side via the transmission antenna 13. On the receiving side, the transmitted signal is received via the receiving antenna 14 and receiver 15 and demodulated to reproduce the baseband signal. Needless to say, this baseband signal is
It corresponds to the multiplexed audio signal itself output from the audio prediction characteristic conversion means 5, and is input to the demodulation processing extraction means 11, and on the other hand, the terminal 11
0 as the original audio signal, and on the other hand, due to the demodulation processing described above,
09 as a predetermined symbol signal. Further, in the embodiment shown in FIG. 2, the case where a wireless line is used as the transmission line has been described, but the scope of application of the present invention is not limited to the case of the above-mentioned wireless line. It also applies to all transmission lines, including optical lines (including spatial lines and optical cable lines).

As described above in detail, the present invention has the advantage that in a communication device intended for transmitting audio signals, predetermined symbol signals can be multiplexed and transmitted in addition to the audio signals.

[Brief explanation of drawings]

1A and 1B are block diagrams showing the main parts of the transmitting side and the receiving side, respectively, in one embodiment of the present invention, FIG. 2 is a block diagram showing the main parts of another embodiment of the invention, and FIG. a and b are explanatory diagrams of audio waveforms before and after the audio prediction characteristic conversion, respectively; FIG. 4 is an explanatory diagram of the correspondence between the multiplexed audio signal and the analysis section in the demodulation processing extraction means; and FIG. 5 is an illustration of the adjacent analysis section. This is an example of spectral distance measurement data between. 1, 6...A-D converter, 2...Modulation processor, 3...
D-A converter, 4...Low pass filter, 5...Speech prediction characteristic conversion means, 7...Window processor, 8...Waveform correction processor, 9...Linear prediction analyzer, 10...Spectral distance measuring device, 11 ...Demodulation processing extraction means, 12...
transmitter, 13...transmission antenna, 14...reception antenna, 15...receiver.

Claims (1)

[Scope of Claims] 1. In a communication device that transmits a predetermined audio signal, in order to multiplex and transmit a predetermined symbol signal to a transmission line of the communication device, the audio signal is converted from analog to digital to digital. an A-D converter that outputs an audio signal; the digital audio signal is inputted, and is converted into a multiplexed digital audio signal in which sampled values are partially omitted via a buffer memory by the control action of the symbol signal; a D-A converter that performs D-A conversion on the multiplexed digital audio signal outputted from the modulation processor and outputs a multiplexed analog audio signal; a low-pass filter that removes high-order frequency components contained in the signal; and a voice prediction characteristic conversion means configured on the transmission side, the multiplexed analog voice signal generated by the voice prediction characteristic conversion means. A-D that converts the received signal into A-D and outputs a multiplexed digital audio signal.
A converter inputs the multiplexed digital audio signal, cuts out and outputs a multiplexed digital audio signal corresponding to an analysis section of a predetermined time length arranged at a time difference of a predetermined sampling period via a buffer memory. A window processor is used to input the multiplexed digital audio signal outputted from the window processor, linear prediction analysis is performed for each analysis interval, and linear prediction coefficients of the multiplexed digital audio signal in each analysis interval are calculated. A linear prediction analyzer that extracts and outputs the linear prediction coefficients, calculates and outputs the spectral distance between the mutually adjacent analysis sections, and also A distance measuring device that detects and outputs the generated discontinuity, a multiplexed digital audio signal outputted from the window processor, and the spectral distance information are inputted, and the time position corresponding to the point where the sampled value drops is referenced. and a waveform correction processor that sends the multiplexed digital audio signal whose waveform has been corrected by performing the waveform correction to the linear prediction analyzer as a target for waveform prediction analysis. A multiplex communication device characterized by: