JPS6324335B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a detection device for modulated data signals. The modulated data signal is obtained by modulating a carrier wave with a digital data signal, and an example is a voice-grade data signal that uses a voice band. This voice-grade data signal is a 4.8 Kb/s or 9.6 Kb/s data signal placed on, for example, a 2 KHz analog carrier wave using a modulation method such as QAM or 4-phase PSK. With the advancement of digital signal processing technology, digital communication is progressing, and in order to make effective use of transmission paths for economical purposes, digital voice insertion method (Digital
Speech Interpolation (DSI) is currently being adopted, and when voice-grade data signals are transmitted over regular telephone lines, the problem arises that DSI frequently freezes out. Originally, the DSI method was based on the fact that during a telephone conversation, it is necessary to listen to what the other party is saying, so the average time each caller actually speaks is about half of the call time.
It has also been developed focusing on the redundancy that there are times in a call that do not need to be transmitted as audio, such as pauses between syllables, words, or phrases, even while speaking. That is, if you instantaneously observe a large number of telephone lines on which calls are being made, you will find that voice signals are present on only less than half of them. Therefore, in the DSI method, the digitized audio signal of each line (input trunk) is divided into unit blocks (several ms in time), which are a predetermined number of fixed sample value groups, and each unit block is divided into A voice detector determines whether or not a voice exists in a unit block, extracts a unit block in which voice exists, and inserts only the extracted unit blocks of each input trunk into each other's empty spaces to create a predetermined number of blocks. It is then reorganized into a DSI output channel and then transmitted. By eliminating the above-mentioned redundancy in this manner, the number of DSI output channels can be reduced to less than half of the total number of input trunks, and the transmission path can be more than doubled. However, when a voice-grade data signal is transmitted over a telephone line instead of a voice signal, the voice-grade data signal is transmitted unilaterally unlike a conversation, and there are no pauses like voice, so the voice detector is It is determined that audio exists in all unit blocks throughout the transmission period. As a result, the number of active input trunks exceeds the number of DSI output channels, causing a freeze-out in which information on any active input trunk is momentarily no longer transmitted. At the time of this freeze-out, the circumstances differ greatly depending on whether the audio signal is momentarily interrupted or when the voice-grade data signal is momentarily interrupted. That is, although a momentary interruption of the audio signal causes some discomfort in terms of audio quality, the receiver can understand most of the content of the call. On the other hand, a momentary interruption of the voice-class data signal renders it useless as information. Therefore, when a voice-grade data signal is transmitted over a regular telephone line, it is necessary to either transmit it on a dedicated line without passing through DSI, or to pass it through DSI without momentary interruption. What is important in this case is to detect as accurately and quickly as possible whether the input signal is a voice-grade data signal or not. This detection can also be done by demodulating the signals of each input trunk of the DSI, but this is not practical because the modulation method of the voice-class data signal cannot be specified. The necessity of detecting voice-grade data signals has been described above in connection with the adoption of the DSI method, but it goes without saying that distinguishing between voice signals and modulated data signals is often required in the field of communications. An object of the present invention is to provide a modulated data signal detection device that meets such demands. The configuration of the present invention to achieve such an object is to determine whether a digitized input signal is a modulated data signal or an audio signal, and includes an adaptive filter and a filter that filters the input signal at regular intervals. a coefficient range determiner for determining whether or not the filter coefficients of each stage of the adaptive filter are within a specific region determined for each stage, for each unit block divided into;
A signal determiner that determines that an input signal is a modulated data signal by determining that the coefficient range determiner exists for a predetermined number or more of consecutive unit blocks. and an adaptive filter for determining whether a digitized input signal is a modulated data signal or an audio signal; a coefficient value region determiner for determining whether or not the filter coefficients of each stage of the adaptive filter are within a specific region determined for each stage; A power calculator that calculates power and compares whether it is equal to or higher than a standard value;
a zero-crossing number counter that counts the number of zero-crossings in each block that divides the input signal into the certain period or a longer period to determine whether or not it is within a specific range; Unit blocks determined to be present by the zero-crossing counter, determined to be equal to or greater than a reference value by the power calculator, and determined to be present by the coefficient range determiner occupy a predetermined number or more of the plurality of consecutive unit blocks. The present invention is characterized in that it includes a signal determiner that uses as a condition for determining an input signal as a modulated data signal. Hereinafter, one embodiment of the present invention will be described with reference to the drawings. In this example, both the audio signal and the modulated data signal are 8KHz, 13bit linear.
It is assumed that the modulated data signal is given as a PCM signal, and the modulated data signal will be explained as a voice-grade data signal with a carrier wave of about 2 KHz. Before explaining the embodiment device, let us touch on various characteristics of the modulated data signal, which are the basis of the present invention, ie, the spectrum, the number of zero crossings, and the power within the block. (a) Since the modulated data signal always passes through a spectrum shaping filter after modulation in order to match the spectrum to the bandwidth of the transmission path, the modulated data signal samples the inherent correlation caused by this shaping filter. The prediction coefficients that have between values and give the best prediction depend on the constants of this shaping filter. The present invention takes advantage of the latter property. (b) The modulated data signal is transmitted over a specific carrier (e.g.
2KHz) and then passes it through the above shaping filter, so the spectrum is almost flat within the band, and unlike an audio signal, the number of zero crossings per unit block can only take values within a certain range based on the above carrier wave. I don't get it. (c) The modulated data signal is typically transmitted with a power of about -12 dBmO, and the power fluctuations are relatively small. Note that the power that has been subjected to 4-phase PSK modulation is almost completely constant, but the power that has been subjected to QAM modulation is subject to amplitude modulation, so there is some fluctuation.
However, the degree of this is very small compared to the audio signal. Detection performance can be improved by utilizing the properties of items (b) and (c) above. In addition, the first
Figures a, b, and c show the power spectra of a 4.8 Kb/s voice-grade data signal, a 9.6 Kb/s voice-grade data signal, and a voice signal, respectively. FIG. 2 is a block diagram showing an embodiment of the present invention. In the figure, 1 is an input terminal, 2 to 5 are output terminals, 6 is a zero-crossing counter, 7 is an adaptive predictor, and 8 is a subtracter. , 9 is a coefficient range determiner, 10 is a power calculator, 11 is a power comparator, 12 is an observation time setter, 13 is a signal determiner, and 16 is an adaptive filter. A digital input signal S is drift-compensated and inputted to the input terminal 1 . The zero-crossing counter 6 receives the digital input signal S by 32
Divide each sample into blocks (4 ms), count the zero crossing number Zc for each block, and when the following formula (1) holds true, set Z-Flag ``1'' as the zero crossing number determination flag for the block, and then Otherwise, Z-Flag "0" is set and sent to the signal judger 13. 7≦Zc≦20 (1) However, the threshold in equation (1) is when the sampling frequency is 8 KHz and the block length is 32 samples. The power calculator 10 divides the digital input signal S into unit blocks (2 ms) of every 16 samples for detailed judgment, and calculates the intra-block power Pw based on the following equation (2) or (3). P-Flag "0", "1", and "2" are set as power classification flags according to the power level classification shown in Table 1, and the signal judgment unit 13, observation time setter 12, and adaptive filter 16 are controlled. The calculated value Pw is sent to the adaptive predictor 7, and the calculated value Pw is also sent to the power comparator 11.
However, in equations (2) and (3), N is the unit block length, in this case N=16, and Xj is the amplitude value of each sample. Pw=1/N _N 〓 ^j=1 ｜Xj｜ …(3)

[Table] The adaptive filter 16 is composed of a subtracter 8 and an adaptive predictor 7, and the details will be described later, but the prediction error e(K) at each sampling point K given by the following equation (4) is squared. The filter coefficients (prediction coefficients) a _i (K) are successively adjusted and controlled so as to minimize the filter coefficients a i (K), and the filter coefficients a _i (K) are sent to the coefficient range determiner 9. However, f(K) is the filter output (predicted value) at each sample point K. e(K)=S(K)-f(K)...(4) In the case of this embodiment, the coefficient range determiner 9 calculates the value for each tap in the adaptive predictor 7 of the adaptive filter 16 (in this embodiment, 4 taps). For the filter coefficients a ₁ (K), a ₂ (K), a ₃ (K), and a ₄ (K) of tap), the intra-block coefficients for each unit block (2 ms) of 16 samples are calculated using the following equation (5). Calculate the average value _i , and set C-Flag “1” as the coefficient judgment flag when all of the following equations (6) hold true.
is set, and in other cases, the C-Flag is set to "0" and sent to the signal determiner 13. _i = 1/N _N 〓 ^j=1 a _i (K+1-j) ...(5) (however, i = 1 to 4, N = 16) However, the threshold of each range in equation (6)
The values of A _T h _1i and A _T h _2i (i=1 to 4) shown in Table 2 are obtained by simulation, and these values are appropriately determined depending on the number of taps of the adaptive filter 16. In addition to making the determination based on the average value, the coefficient range determiner 9 determines C- for the unit block when the individual filter coefficient _a Flag “1” may be set.

[Table] The signal determiner 13 uses the above-mentioned Z-Flag, P-
Flag and C-Flag are input, and the temporary detection flag d-Flag "1", which indicates the possibility that there is data within a unit block of 16 samples (2 ms), is set by the judgment logic of the following equation (7). Specifically, this d-
Data determination flag D-Flag that determines that the input signal S is a voice-grade data signal when Flag “1” stands three or more times within a continuous 4-unit block (8ms)
It sets "1" and sends it to output terminal 2. However, the conditions for setting D-Flag "1" based on d-Flag "1" and "0" are not limited to the above-mentioned [3 or more times/4 unit block], but can be determined as appropriate. The signal determiner 13 of this embodiment further has the following functions upon determination. The first function is to conclude the data detection judgment in a short time, and is controlled by the output t of the observation time setting device 12. This observation time setting device 12 is a time counter,
For example, 20ms after P-Flag “2” is set once,
That is, it counts up to 10 unit blocks, and does not count when this is exceeded. Therefore, P-Flag “2”
If the condition for establishing D-Flag “1” is not satisfied within 20ms after P-Flag is set, at least P-
Output terminal 3 until Flag becomes “0” or “1”
The V-Flag is set to "1", and the input signal S during that time is regarded as an audio signal, the determination is terminated, and the next determination is awaited. The second function is to improve the accuracy of data detection, and is controlled by Z-Flag "0". That is, if Z-Flag "0" exists in at least one block (4 ms) within 20 ms of the rise of P-Flag "2" mentioned above, D-Flag
Even if the conditions for establishing Flag “1” are met, this is ignored, and from that block onwards, V- is applied to the output terminal 3 at least until P-Flag becomes “0” or “1”.
Flag “1” is set and during that time it is regarded as an audio signal to avoid the risk of misjudgment. The third function is D-Flag
It identifies the modem content of the voice class data signal when "1" is set, and uses the output R of the power comparator 11 as information. The power comparator 11 calculates the intra-block power Pw from the power calculator 10 and the power comparator 11 according to the above equation (2).
Alternatively, the ratio R=20logPw/ _PwA of the prediction error e calculated according to the above equation (3) to the intra-block power _PwA is calculated and sent to the signal determiner 13. Signal judger 13
is D- for 20ms after P-Flag “2” rises.
Calculate the average value of the ratio R = 1/K _K 〓 ^{L = 1} R _L only for the unit block where Flag “1” is set, and when ≧6.0, it is assumed that the modem data is 4.8 Kb/s or less, and the output terminal Set M ₁ −Flag “1” to 4, and conversely set <6.0
At this time, it is assumed that the data is 9.6 Kb/s modem data, and M ₂ -Flag "1" is set at the output terminal 5. As can be seen from Figure 1 a and b, this identification condition is
This is because the 9.6 Kb/s voice-grade data signal (Figure 1b) has a wider flat part than the 4.8 Kb/s voice-grade data signal (Figure 1a), so the prediction error is correspondingly larger. It's on. Now, the aforementioned adaptive filter 16 will be explained in detail with reference to FIG. 3. As mentioned above, the adaptive filter 16 is composed of the subtracter 8 and the adaptive predictor 7, and the adaptive predictor 7 has a delay time corresponding to one sample of the input signal S (1/1 in the embodiment). 8000s = 0.125ms) delay devices DD ₁ , DD ₂ , DD ₃ , DD ₄ are connected in series to each tap output, and weight coefficients (filter coefficients) a ₁ , a ₂ , a ₃ are applied to each tap output of this series circuit. , a variable coefficient multiplier that multiplies by ₄ and outputs w ₁ , w ₂ , w ₃ , w ₄
, an adder 14 that sums up the outputs of each of the coefficient multipliers w ₁ , w ₂ , w ₃ , and w ₄ and outputs a filter output (predicted value) f to the subtracter 8; and each of the coefficient multipliers w ₁ , _w2 ,
A coefficient controller 15 is provided to control filter coefficients _{a 1 ,} a ₂ , a ₃ , and a 4 of w ₃ and w ₄ to optimally predict. This coefficient controller 15 controls i=1 to 4 so as to minimize the square of the prediction error e(K) at each sampling point K.
The filter coefficient a _i (K) of is sequentially adjusted and controlled. The coefficient controller 15 in this embodiment uses the steepest descent method (maximum slope method), and calculates the filter coefficient a _i (K) at the current moment K and the input signal sample value S ( K) and prediction error e(K), and past input signal sample values S(K-4) to S(K
-1) and determine the filter coefficient a _i (K+1) at the next time point (K+1). In equation (8), g is a constant that can also be called a sensitivity constant, and if the constant g is small, a _i (K) → a _i (K+1)
If the change in is slow and the constant g is large, a _i (K)→a _i
The change in (K+1) is steep. In this example, in order to quickly converge and stabilize the prediction, this constant g is set to 0.00 based on the contents of P-Flag and d-Flag.
It is available as a three-value variable of 0.05 and 1.00. Table 3 shows the selection conditions for this constant g. However, g=0.00 means no operation.

【table】

[Table] Incidentally, the settings of the initial value of the filter coefficient of the adaptive predictor 7 and the initial output of the delay device are not particularly limited, but in this embodiment, both are set to 0.00.
As is clear from the selection of constant g in Table 3, P-Flag changes from “0” to “1” or “0”.
→ When the transition is “2”, g=1.00, and the coefficient of the adaptive predictor 7 starts from the initial value 0.00. After the start, the constant g is changed based on Table 3, and each time the d-Flag is set to 1, the initial value of the filter coefficient is set to the average of all blocks for the next block (16 samples, 2 ms). It is set to the value and restarted. This is d-
This is because the input signal S has a possibility of being data when Flag "1" is set, and the purpose is to quickly converge the prediction. The initial value of the filter coefficient for the next block described above is determined by giving the average value of the filter coefficients in the block where d-Flag is set to "1", and is determined by the following equation (9) using the previous equation (5). . a _i (K+1)= _i =1/N _K 〓 ^L=1 a _i (K+1-j) ...(9) (However, N=16, i=1 to 4) By the way, in the above embodiment, the modulated data signal The configuration of the detection device includes an adaptive filter 16, a coefficient range determiner 9, and a signal determiner 13.
Although the power calculator 10 and the zero-crossing counter 6 have been described above, it is possible to exclude the power calculator 10 and the zero-crossing counter 6 from these as the basic configuration of the device of the present invention. . In other words, basically adaptive filter 1
6, a coefficient range determiner 9 that sets a coefficient determination flag C-Flag "1" when the filter coefficient a _i of each tap of the adaptive filter 16 is in a specific region, and a coefficient determination flag C-Flag " A modulated data signal detection device can be configured with a signal determiner 13 that determines whether or not the input signal S is a modulated data signal based on the continuity of occurrences of 1''. The reason is that the modulated data signal is passed through a spectrum shaping filter, so the input signal S
This is because when is a modulated data signal, the filter coefficient a _i for each tap of the adaptive filter 16 takes a value in a specific region depending on the constant of the spectrum shaping filter. Therefore, power calculator 1
Even if the 0 and zero crossing counter 6 are not necessarily used together, the coefficient judgment flag C-Flag
Among multiple unit blocks with consecutive “1”s,
If it occurs in a predetermined number or more unit blocks (for example, three or more times in four unit blocks), it can be determined that it is a modulated data signal. Furthermore, for this basic configuration, as shown in Figure 2,
When the zero crossing number Zc for each unit block of the input signal S is within a specific range, the zero crossing number judgment flag Z-
The zero-crossing counter 6 sets Flag "1", calculates the intra-block power Pw for each unit block of the input signal S, and sets the power classification flag P-Flag "0", "1",
By adding the power calculator 10 that sets "2", the signal determiner 13 temporarily detects that the logic product of Z-Flag "1", P-Flag "2", and C-Flag "1" stands. Based on the continuity of occurrence of the flag d-Flag "1", it is determined whether the input signal S is a modulated data signal, and error detection in data signal determination can be improved. In addition, in the embodiment, the operation (start/stop, selection of constant g, etc.) of the coefficient controller 15 of the adaptive predictor 7 is controlled by the contents of P-Flag and d-Flag. This speeds up prediction convergence and helps in early judgment. Furthermore, Z-
If Flag “0” is adopted with priority over data signal negation, false judgments will be eliminated. Although the block length for zero-crossing number Zc counting is longer than the other unit blocks, this is to improve counting accuracy, and both the block length and unit block length may be determined as appropriate.
In addition, the power classification flag P-Flag is “0”,
The three values are “1” and “2”, but this is also “0” for example.
It may also be set to a binary value such as ``1'' and ``1'' together. Furthermore, it goes without saying that the various thresholds and judgment conditions are not limited to the numerical values or criteria shown in the embodiments. Further, a Kalman filter may be used instead of the adaptive predictor 7 forming the adaptive filter 16. Furthermore, similar to the audio detector, a hangover time may be provided in which the modulated data signal continues even if detection is interrupted within a certain period of time after the modulated data signal is detected. As specifically explained above with the examples,
According to the present invention, since an adaptive filter is used with focus on a spectrum shaping filter, it is possible to detect whether an input signal is a modulated data signal regardless of the modulation method. In particular, when equipped with a power calculator and a zero-crossing counter, for example 20ms after the start of operation.
It is possible to detect whether the signal is a modulated data signal or not in such a short time. In addition to the above-mentioned applications in the DSI system, the device of the present invention is also very useful in the field of digital communications for distinguishing modulated data signals from voice signals.

[Brief explanation of drawings]

Figures 1a, b, and c are diagrams showing the power spectra of various signals, Figure 2 is a block diagram showing an embodiment of the present invention, and Figure 3 is a block diagram showing the main parts in Figure 2 in detail. It is. In the drawing, 6 is a zero-crossing number counter, 7 is an adaptive predictor, 8 is a subtracter, 9 is a coefficient range determiner, 1
0 is a power calculator, 11 is a power comparator, 12 is an observation time setter, 13 is a signal judger, 14 is an adder,
15 is a coefficient controller, 16 is an adaptive filter, DD ₁ to _{DD 4} are delay devices, and w ₁ to w ₄ are coefficient multipliers.

Claims (1)

[Scope of Claims] 1. A device for determining whether a digitized input signal is a modulated data signal or an audio signal, comprising an adaptive filter and a unit block that divides the input signal into predetermined intervals. For each stage, a coefficient range determiner determines whether or not the filter coefficients of each stage of the adaptive filter are within a specific region determined for each stage, and a determination that the coefficient range determiner exists is performed continuously. What is done to a predetermined number or more of the plurality of unit blocks to be
1. A detection device for a modulated data signal, comprising: a signal determiner that uses an input signal as a condition for determining a modulated data signal. 2. In a device that determines whether a digitized input signal is a modulated data signal or an audio signal, an adaptive filter and an adaptive filter are used for each unit block in which the input signal is divided into fixed intervals. a coefficient range determiner for determining whether the filter coefficients of each stage of the filter are within a specific range determined for each stage; and a coefficient value range determiner for calculating the power per unit block of the input signal to determine whether the power per unit block of the input signal is greater than or equal to a reference value. a power calculator that compares whether or not the input signal is within a specific range; A zero-crossing number counter that determines the number of zero-crossings, and a plurality of consecutive unit blocks that are determined to be present by the zero-crossing number counter, are determined to be present by the power calculator, are determined to be present by the power calculator, and are determined by the coefficient range determiner. 1. A detection device for a modulated data signal, comprising: a signal determiner that determines that an input signal is a modulated data signal by determining that the input signal occupies a predetermined number or more of unit blocks.