JPH0865717A - Dial detection device - Google Patents

Abstract

PURPOSE: To provide a dial detection device which can accurately detect a dial pulse signal out of the reception signal sent from a telephone circuit. CONSTITUTION: The 1st formant of a reception signal is attenuated by an HPF at a preprocessing part 5 and the absolute value of the formant is calculated and shifted to the DC side. Then the voice signals are deleted by the thinning out the processing so that a low-speed sample signal 27 is outputted. The mute pulses and abnormal waveforms are eliminated at a head digit dial detection part 6 so that a head digit dial 28 is obtained. The termination of a dial pulse signal is detected at a subsequent digit dial detection part 7 so that a subsequent digit dial 29 is obtained with no influence of mute pulses. Then a PB signal 30 is transmitted. Thus the accurate dial detection is attained even when the head digit dial is not known yet and also the voice signals are superposed on each other.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a dial detecting device for detecting a harmonic component of a pulse type dial signal transmitted through a public telephone line and converting it into dial information. More specifically, when the dial key of the telephone is operated to transmit some information to the callee after the telephone line is connected to the callee, the device at the callee must be sure to detect the dial information. The present invention is intended to provide a new dial detection device that can be used.

[0002]

2. Description of the Related Art A center device connected to a public telephone line is provided with a voice guidance device and a dial signal receiving device, and when a telephone call arrives from the public telephone line side, voice guidance is provided to prompt input of a member number and a product number. There is a mail order service in which members enter numerical data with dial keys to confirm the product and receive an order. There is also a service similar to this in which a ticket is reserved or financial information is distributed. Further, remote control such as controlling and reproducing the contents stored in the answering machine from a telephone at a remote place, or dial-in for selecting an extension telephone of PBX is widely used.

Numerical data input for performing these controls generally uses dial keys provided in a telephone. This is because inexpensive and general-purpose input can be performed by using a key attached to the telephone as an input means without using a dedicated input means.

P as means for transmitting dial key information
There are B (push button) signal system and pulse system.
The PB signaling method is more convenient for realizing services and the like that use dial key information. This is because information can be reliably transmitted using the frequency band allowed for the telephone line without being affected by the direct current limitation in the relay circuit of the public line. Since the original purpose of dialing is to inform the local subscriber exchange of the destination number, any method is the same as far as this original purpose is concerned. Have joined.

A service using dial key information is difficult to obtain an economic effect without a wide range of members. It is expected that a wide range of members will participate if dial key information can be entered from a pulse-type subscriber line telephone, but the direct current in the circuit of the relay system is cut and the frequency band 300
Since there is a direct current limitation of passing a telephone signal of up to 3.4 kHz, it is not easy to detect a dial pulse that turns on / off direct current, and the range of use of dial key information has to be limited.

The reason why it is not easy to detect the dial pulse waveform having no DC component after passing through the relay system is as follows.

First, the dial pulse waveform has a speed of 1
There is no feature to distinguish from a voice signal and line noise except that the make rate is 30% at 0 pps or 20 pps.

Second, since the dial pulse waveform representing the dial "1" or dial "2" is a single-shot waveform or a waveform close to it, it cannot be detected by a tuning filter or the like characterized by periodicity.

Thirdly, the reception level of the dial pulse waveform transmitted through the telephone line fluctuates greatly from -10 dBm to -40 dBm, but gain control is performed when the received signal is unclear whether it is a voice waveform or a dial pulse waveform. Stable detection is not possible because it cannot be applied.

Fourth, since the rotary dial type telephone performs dial mute (to prevent the dialer's handset from making annoying loud dial tone), the mute start time and end time, that is, Dial ・ Before and after the pulse waveform,
Since a pulse waveform very similar to the pulse waveform is generated, it is easy to make an erroneous detection.

Fifth, the voice signal contains a signal similar to the dial pulse waveform, although it is rare, and may be erroneously recognized as the dial pulse waveform representing the dial "1".

Sixth, since the dial pulse waveform has various waveform shapes depending on the combination of the telephone set, the local subscriber exchange, and the relay system, it is impossible to predetermine the waveform shapes and distinguish the waveforms.

There is a method of learning the waveform as a relatively effective detection method of the dial pulse waveform for solving such various problems. This is a method in which the dial pulse waveform is detected in some way during the initial period of dial reception, the characteristics of the waveform at that time are stored, and thereafter the characteristics are used for recognition. According to this method, many characteristics relating to the waveform, including the waveform amplitude, can be used for recognition, so that the recognition rate can be improved not only for recognition of the dial pulse waveform having periodicity but also for recognition of the dial "1". However, this method requires a large amount of memory to store the characteristics of the dial pulse waveform.
There is a problem in that the amount of processing required for comparative recognition with the information representing the characteristics of the pulse waveform is large, making recognition in real time difficult. Further, if the detection of the dial pulse waveform is unsuccessful in the initial period, there is a risk that the subsequent recognition cannot be performed at all.

In Japanese Patent Laid-Open No. 2-231856, in order to reduce this risk, dials "0" and "2", which are easy to detect for a long duration as initial learning dial pulse waveforms, are sampled pulse signal trains. Was used as. Since the receiving side knows the numerical value in advance, the waveform information such as the period, the receiving level or the presence of the mute pulse is collected using the time template or the amplitude template. However, limiting the first dial to a known value on the receiving side imposes a limit on the degree of freedom of the numbering plan for various services and the like, and therefore an improvement is desired.
Also, in this type of device, information was passed to the center device on the assumption that the recognition rate was as close as possible to 100%, but the actual telephone line exhibited extremely many transfer characteristics and exceeded the prediction range. May receive shaped dial pulse waveforms. Therefore, when the recognition limit is exceeded, it is required to notify the center device.

Further, in a usage mode in which the center device prompts the user on the public telephone line side to dial by voice guidance, the voice guidance signal is wraparound dial / pulse waveform in the 2-wire / 4-wire conversion section in the dial detection device. Because of the superposition, the dial pulse could not be detected during voice guidance. Therefore, the center device tells the user to dial after the confirmation sound in advance in the voice guidance, and outputs the confirmation sound at the end of the voice guidance to detect the dial pulse during the silent period. However, since the content of the voice guidance is known in advance to the person who uses it many times, he / she has to wait until he actually hears the confirmation sound before dialing.

[0016]

When the problems to be solved by the present invention are listed, the first is to reliably detect the first dial pulse waveform after the start of communication, and the second is to reliably detect the first dial pulse waveform after the start of communication. The first is to detect even if the numerical value of the dial is unknown, the third is to reduce the memory and processing amount by reducing the dial pulse waveform information to be stored, and the fourth is the first digit. Is to reliably recognize the isolated dial pulse waveform in the subsequent digits following
The fifth is to eliminate the dial pulse waveform even if it is accompanied by the dial / mute waveform and perform accurate recognition, and the sixth is to make the recognition uncertain depending on the state of the telephone line.・ To detect the pulse waveform and notify the center device that is the recipient of the dial signal,
The seventh is to take a defensive measure so as not to erroneously recognize the PB signal or the voice signal, and the eighth is to perform dial detection independent of the reception level. The ninth is to perform dial detection even during voice guidance.

[0017]

SUMMARY OF THE INVENTION A device according to the present invention is a high-pass filter which does not shape the waveform of a voice and attenuates a component corresponding to the first formant, and a high-pass filter output is rectified to direct a signal band to a direct current. By an absolute value calculation unit that shifts to the side, a decimation processing unit that reduces the number of samples, and a preprocessing unit that removes the voice superimposed on the dial pulse waveform, and a difference calculation that does not depend on the reception level. Check the periodicity to detect that it is a dial pulse waveform, calculate the change around the average value to eliminate the PB signal, etc., and determine the period and tip and end of the waveform to determine the dial numerical value. A leading digit dial detection unit that recognizes and informs of an abnormal waveform, and further finds and stores the waveform and characteristics of the waveform, and a leading digit dial detection unit It detects the dial pulse waveform of the succeeding digit based on the stored dial pulse waveform and waveform information, removes the mute waveform, etc. from the detected dial pulse waveform, and recognizes the dial digit value. The main part is constituted by a PB signal transmitting section for converting the recognized dial numerical value or abnormal waveform information into a PB signal and transmitting it to a system such as a voice guidance device in the subsequent stage.

[0018]

The operation of the device according to the present invention will be listed below.

First, the frequency component corresponding to the first formant, which is the main energy component of voice, is attenuated by the high-pass filter, and the periodicity of the dial pulse waveform is selectively detected by the difference calculation. The dial pulse waveform can be detected without erroneously detecting voice or line noise.

Secondly, since the waveform selection by the difference calculation can be performed if the dial pulse waveform is repeated a certain number or more, the dial pulse waveform can be detected even if the dial digit value of the leading digit is not known.

Thirdly, since the frequency component of the dial pulse waveform is shifted from the telephone band to a band near DC by rectifying it in the preprocessing section, and the sampling frequency is lowered by the thinning process, the dial pulse waveform is stored. Can be realized with a small amount of memory and can be processed in real time.

Fourth, since the succeeding digit is detected using the leading digit dial pulse waveform and the waveform information, the dial pulse waveform of the succeeding digit is not required to have periodicity, and the isolated dial pulse waveform Can be detected.

Fifth, since the exclusion judgment of the mute waveform is made in the recognition of the first digit dial or the recognition of the subsequent digit dial, it is possible to accurately recognize the dial numerical value.

Sixth, since the abnormal waveform is judged after the detection of the leading digit dial pulse waveform, it is possible to prevent an incorrect recognition result from being output.

Seventh, the leading digit dial detector determines the amount of change in the value compared with the average value of the waveform and dials the succeeding digit dial.
In the detection of the pulse waveform, the detection is made based on the waveform information stored in the leading digit dial detecting section.
It is possible to prevent erroneous detection of the B signal or the voice signal.

Eighth, by normalizing the difference calculation, it becomes possible to detect the dial pulse waveform independent of the reception level.

Ninth, since the two-line / four-line converting unit in the dial detecting device is adapted to remove the circling voice by the voice guidance, the dial pulse waveform can be detected even during the voice guidance.

[0028]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows the circuit configuration of an embodiment of the present invention. Reference numeral 21 is a telephone line for transmitting a dial signal or a voice signal, and 1 is a two-line 4 for closing a DC loop of the telephone line and for separating a transmission signal 23 and a reception signal 22.
A telephone line interface for line conversion, 2 is an AD conversion circuit for sampling an analog signal at 8 kHz and converting it into a digital signal, 3 is a DA conversion circuit for converting an 8 kHz sampling signal into an analog signal, 4
Is an echo replica created from the transmission signal 25 by digital processing in order to reduce the echo, that is, the sneak of the transmission signal 23 into the reception signal 22 caused by the imbalance of the 2-line to 4-line conversion unit in the telephone line interface 1. The echo canceller cancels the echo component contained in the reception signal 24 and outputs it as the reception signal 26.

Reference numeral 5 is a dial pulse waveform pre-processing unit for performing high-pass filter processing, reduction of the number of calculations, and thinning-out processing for reducing information to be stored.
Sample signal for ps dial pulse waveform detection and 10
2 of sample signal for pps dial pulse waveform detection
6 is a low-speed sample signal composed of two signals, and 6 is a leading digit dial detection unit that detects the first dial pulse waveform immediately after the start of communication, recognizes the dial pulse number, extracts the waveform feature, and stores the waveform. Reference numeral 7 is a succeeding digit dial detecting unit that detects the dial pulse of the second digit and thereafter using the information obtained by the leading digit dial detecting unit 6,
28 is the leading digit dial numerical information recognized by the leading digit dial detecting unit 6, 29 is the succeeding digit dial numerical information recognized by the succeeding digit dial detecting unit 7, and 8 is the leading digit dial numerical value and the succeeding digit dial. Numerical information 28, 29
Is a PB signal transmission unit for converting the PB signal into a dial signal 30.

Numeral 9 indicates that the reception signal is stopped when the PB signal is transmitted.
A switch (SW) that outputs a B signal, and 10 is a DA converter that converts a reception signal or a PB signal from an 8 kHz sampling signal into an analog reception signal 31.
Reference numeral 11 is an AD converter for sampling the analog transmission signal 32 at 8 kHz and converting it into a digital signal, reference numeral 33 is a system signal line equivalent to a telephone line connecting with a system that provides services using a dial signal, and 12 is A system line interface for converting the transmission signal 32 and the reception signal 31 into four lines and two lines, supplying DC to the system, and detecting a DC loop closure, and 34 is DC loop closure information in the system line interface 12. The DC loop closure information 34 also detects the dial pulse waveform after initializing the signal processing unit 5, the leading digit dial detection unit 6, the succeeding digit dial detection unit 7, and the PB signal transmission unit 8 when the DC loop is closed. It is also used as control information for stopping dial pulse detection when the DC loop is opened.

In FIG. 1, the echo canceller 4, the preprocessing unit 5 and the PB signal transmitting unit 8 are provided in the telephone band (300-).
Since it is a signal processing of 3.4 kHz), it is necessary to process the signal at a sampling rate of 8 kHz. However, the leading digit dial detecting unit 6 and the succeeding digit dial detecting unit 7 reduce the speed of the low speed sample signal 27 by the preprocessing unit 5. 8 because it is processing
It is inefficient to process the program mixed with the kHz sample signal. Therefore, in one embodiment of the present invention, a DSP (digital signal processor) that realizes the functions of the preprocessing unit 5, the leading digit dial detecting unit 6, the succeeding digit dial detecting unit 7, and the PB signal sending unit 8 according to a program is installed at 8k.
It is divided into two types, one for Hz samples and the other for low speed samples.

In the 8 kHz sample DSP, one sample 125 μs is time-divided to perform echo canceller signal processing,
Preprocess dial pulse waveform and generate PB signal. The low-speed sample DSP divides the processing sequence into the first digit dial detecting step and the subsequent digit dial detecting step. The low-speed sample signal is input / output between the two DSPs by serial transfer, and the leading and subsequent digits dial numerical information 28 and 29, which are dial recognition results, are transferred by 8-bit parallel transfer. Although the embodiment of the present invention has been described based on an embodiment in which it is realized by using two DSPs as described above, a DS capable of high-speed arithmetic processing is provided.
If P is used, it is not necessary to separate it into two.

In one embodiment of the present invention, the RAM is repeatedly used so that the processing can be realized by a DSP having a 512-word data RAM. In FIG. 1, the leading and subsequent digits dial numerical information 28, 29 after recognition of the dial numerical value is converted into a PB signal and transmitted to the system through the system signal line 33. Information 28, 29
May be communicated to the system using an interface such as RS232C.

FIG. 2 shows a detailed circuit configuration of the preprocessing section 5. Reference numeral 26 is a received signal that is sampled at 8 kHz and the echo is removed. Reference numeral 41 is a high-pass filter (HPF) for the purpose of attenuating a first formant having a large electric power in a voice signal which becomes an interference signal in performing dial / pulse detection, and further, a low-frequency cutoff of a telephone line. It also serves the purpose of removing ringing caused by group delay distortion near the frequency (about 300 Hz).

FIG. 3 shows a waveform diagram for explaining the operation of the high-pass filter 41 and the absolute value calculator 42. This figure shows a dial pulse waveform generated before and after the make-up period of the dial pulse due to band limitation. The waveform of the received signal 26 is shown in (a1) and the waveform shown in (a1) is shown in (a2). When the absolute value is directly obtained from the signal shown in (b1) of FIG.
The waveform obtained after passing through 1 is shown in (b1) as (b1)
The case where the absolute value is calculated by the absolute value calculation unit 42 from the waveform shown in FIG. There is ringing in the figure (a
1) In the case of (a2), it is equivalent to passing through a transmission system with a slow response, and it is difficult to form a boundary between the pulses, but in the cases of (b1) and (b2) in which the ringing is suppressed, the pulse is suppressed. The boundary between them becomes clear.

The cutoff frequency of the high-pass filter 41 is preferably selected to be about 600 Hz for the purpose of attenuating the first formant of the voice signal and the components near the low cutoff frequency (300 Hz) of the telephone line 21. The phase characteristic of the high pass filter 41 must be linear in order to prevent group delay distortion from being generated in the cutoff region of the pass filter 41. A filter having a linear phase characteristic can be easily realized by symmetrically selecting the tap coefficient of the FIR filter. The absolute value calculation unit 42 shifts the signal band to a low band including direct current in order to shift the occupied band of the received signal to the low band side, which enables the thinning-out process described below.

Reference numerals 43a and 43b are low-pass filters (LPF) required before thinning, and 43a is 20
pps for dial pulse waveform, and 43b for 10pps dial pulse waveform. Low pass filter 43
The cutoff frequencies of a and b will be described. The signal after absolute value calculation has the same cycle as the basic speed of dial pulse (20 pps or 10 pps), but 20 pps and 10 p
In the ps dial pulse waveform, the make period and the break period may be different, or a pulse waveform may occur in the middle part of the break period or the make period due to a waveform clip or the like due to the subscriber circuit on the sender side of the telephone line. Therefore, the frequency band needs to be about five times the basic period.

Therefore, the low pass filter 4 has a frequency of about 100 Hz for a 20 pps dial pulse waveform and a frequency of about 50 Hz for a 10 pps dial pulse waveform.
The cutoff frequencies are 3a and 3b. In the embodiment of the present invention, 20
Low pass filter 43 for pps and 10 pps
The cutoff frequencies of a and b are selected to be 125 Hz and 62.5 Hz. Since the low-pass filters 43a and 43b must have a linear phase characteristic like the high-pass filter 41, they are constructed by symmetrical FIR filters.

Reference numerals 44a and 44b denote thinning-out processing units for performing thinning-out processing to reduce the sampling speed.
0pps for dial pulse waveform, 44b for 10pps
For dial pulse waveform. The sample rate after thinning will be described. The sampling theorem teaches that the sampling frequency after decimation should be at least twice the cut-off frequency of the low-pass filters 43a and 43b. However, in the embodiment of the present invention, dialing is performed by calculating the difference between pulse waveforms. -It is necessary to detect the basic velocity (cycle) of the pulse, and time resolution becomes a problem. It is equal to the sample period, and the sample frequency is increased to improve the resolution.

In the embodiment of the present invention, the sample frequency after thinning is selected to be four times the cutoff frequency of the low-pass filters 43a and 43b in order to make the sample frequency as low as possible and ensure the resolution. That is, in the thinning processing unit 44a, 500 Hz
1 from 8 kHz sample signal to get sample signal
One sample is extracted every 6 samples and the low-speed sample signal 27a is output. Also, in the thinning processing unit 44b, 250H
To obtain the z sample signal, one sample is extracted every 32 samples from the 8 kHz sample signal and the low speed sample signal 27b is output. As described above, in the present invention, the sampling frequency is set to 20 by the absolute value calculation and the thinning processing.
1/16 at 8 kHz for pps, 1 / for 10 pps
The number is reduced to 32 to significantly reduce the calculation amount and the storage amount in the subsequent processing.

Reference numerals 45a and 45b denote voice removing units, which remove voice signals superposed due to wraparound in the 2-line to 4-line converting unit included in the telephone line interface 1 that interferes with dial / pulse detection. To do. Voice pulse guidance from the system is required to detect dial / pulse during voice guidance, for example, when the voice guidance is sent from the system through the system signal line 33 to prompt the telephone user to dial.

At this time, the voice signal that cannot be completely removed by the echo canceller 4 and spills from the transmission signal 25 side to the reception signal 26 side is superimposed on the dial pulse waveform and is dialed.
It interferes with pulse detection. The low-speed sample signals 46a and 46b, which are the inputs to the voice removing units 45a and 45b, have already taken absolute values by the absolute value calculation unit 42 and have been passed through the low-pass filter (LP
F) Since it passes through 43a and 43b, it has a waveform representing electric power. Since the average power of the voice signal changes more slowly with time than the average power of the dial pulse signal, this feature can be used to remove only the voice signal power part through a high-pass filter for the low-speed sample signal. it can.

FIG. 4 and FIG. 5 show waveforms for explaining the operation of one of the sound removing units 45a, 45b and the sound removing units 45a, 45b which can be shown by the same circuit structure. .

FIG. 5A shows a dial pulse waveform which is a low-speed sample signal 46 which is an input signal of the voice removing unit 45 and which is superimposed on the power of the voice signal which changes gently due to wraparound.

The high-pass filter (HPF) 51 removes the low-frequency component, which is the power of the audio signal that changes gently, and outputs the signal 54 of FIG. 5B. The cutoff frequency of the high pass filter (HPF) 51 is 20 p for the purpose of passing the dial pulse waveform and attenuating the voice power waveform.
It is preferable to select a cutoff frequency of about 10 Hz in the case of ps and 5 Hz in the case of 10 pps, and since the cutoff frequency must have a linear phase characteristic like the high pass filter (HPF) 41, a symmetrical FIR filter is used. However, by passing a high-pass filter (HPF) 51,
The waveform of the signal 54 in (b) is shifted to the negative side by the dial pulse average power indicated by the broken line because the average power of the dial pulse is also removed. DP average power adder 52
Then, the signal 54 from the high pass filter (HPF) 51
From this, the average power of the dial pulse shown by the broken line in FIG. 5 (b) is estimated and added.

Since the signal 54 from the high pass filter (HPF) 51 is a dial pulse power waveform,
The portion swinging to the negative side as shown by the broken line in FIG. 5 (b) cancels the average power of the dial pulse, and the envelope on the negative side shown by the wavy line is the negative waveform of the average power of the dial pulse. Equivalent to. The DP power adding section 52 detects the negative-side envelope waveform shown by the broken line, adds it to the signal 54, and raises it to obtain the waveform of the signal 55 in FIG. 5C. The detection of the envelope waveform is performed by, for example, a high pass filter (HPF) 5
It is obtained by clipping the positive side signal of the output side signal 54 of 1 to 0 (half-wave rectification) and passing it through a moving average filter.

Since the sample which has a negative value in the signal 55 after the addition of the average power of the dial pulse is an unnecessary part of the power waveform, it is clipped to 0 by the limiter 53 and the sample is shown in FIG.
The low-speed sample signal 27 of (d) is obtained.

FIGS. 6 and 7 show the flow of signal processing until the presence of the dial pulse waveform is detected in the leading digit dial detection section 6. When the leading digit dial detection operation starts, one sample of the low-speed sample signal 27a is input at a cycle of 2 ms (S1a, FIG. 6), or
One sample of the low-speed sample signal 27b is input at a cycle of 4 ms (S1b). Although not shown, input samples are sequentially stored in a FIFO (First In First Out) format RAM. The RAM is available for 20 pps and 10 pps, and is configured to store waveforms of about 8 basic cycles, that is, 200 samples, respectively. Next, it is determined whether the signal power, that is, the variation V (n) in the vicinity of the average value A (n) shown in the equation (3) described later is a predetermined value, for example, -40 dBm or more (S2Y). If the signal power is extremely low, it is highly likely that the dial pulse waveform of the other line due to crosstalk is present, and V (n) in the denominator of equation (2) described later is used.
This is because that is close to 0 and causes an abnormality in the calculation. If the signal power is small, wait for the next low-speed sample signal (S
2N).

Next, it is determined whether the amplitude change is greater than or equal to a predetermined fluctuation width (S3Y). Signals such as PB dial signals and facsimile procedure signals have a constant waveform envelope,
This is because the conditions of the cycle and the power will be satisfied if the calculation is performed using the equations (2) to (4) described later. One of the objects of the present invention is to pass a PB dial signal or the like having a constant waveform envelope without being touched (S3N), and to detect only a dial pulse waveform whose amplitude change is a predetermined value or more. For example, the amplitude change can be evaluated by the following expression (1) using the change amount V (n) calculated by the expression (3) and the average value A (n) calculated by the expression (4). In the equation, choose α to be about 0.3. V (n) / A (n) ≧ α (1)

Next, a difference calculation is performed (S4). By normalizing the difference calculation, the presence of the dial pulse waveform can be detected regardless of the level of the received signal 26. In the embodiment of the present invention, 5 cycles are prepared for 20 pps and 10 pps, and the difference calculation is performed. The reason for setting five is because the dial speed varies within a range of ± 10%. For example, when the speed is 20 pps, a period of 25 samples (50 ms) is applied to detect the basic velocity (cycle), and the short cycle side is set. 24 sample periods (48 ms) and 23 sample periods (46 ms) are applied, and 26 sample periods (52 ms) and 27 sample periods (54 ms) are applied to the long period side.

Cycles that do not match these (for example, 49 m
In principle, the difference calculation result in the case of a dial pulse having an intermediate period of s, 51 ms, etc. does not become zero. Since the deterioration of the difference calculation result in the case of dial pulse input with an intermediate cycle depends on the sampling frequency after thinning, this relationship will be described in more detail. When the sampling frequency after thinning is set to be twice the cutoff frequency of the low pass filters (LPF) 43a and 43b, the time resolution in the difference calculation is 4 ms, similarly, when it is 4 times, it is 2 ms, and when it is 8 times, it is 1 m.
s. When doubled, the middle of the time resolution (for example, 4
The difference calculation result (r (n) obtained by the equation (2)) is 0.6 when dial pulse is input with a cycle of 8 ms, 52 ms, etc., and similarly, when it is quadruple, it is an intermediate time resolution (for example, 49 m).
s, 51 ms, etc., the difference result is 0.3, and when it is 8 times, the middle of the time resolution (for example, 49.5 ms, 50.5 ms).
Etc.) will be 0.15.

If the sampling frequency after thinning is increased and the time resolution is made fine as described above, a good result can be obtained in the difference calculation, but at the same time, the calculation amount in the processing also increases. On the other hand, when voice is actually input and difference calculation is performed,
A difference calculation result of about 0.4 may be obtained regardless of the sampling frequency after thinning. In this embodiment, the sampling frequency is set to four times the cutoff frequency of the low pass filters (LPFs) 43a and 43b from the viewpoint of avoiding erroneous detection due to voice and reducing the amount of calculation or storage.

In the difference calculation, the difference at the sampling time n indicating the temporal sampling position is r (n), the detected cycle is T, the low-speed sample signal is X (n), and the variation near the average value is V ( n) and the average value is A (n), and the calculation is performed as in equations (2) to (4). A well-known correlation function may be used to detect the periodicity. However, in the case of a waveform that contains a large amount of direct current component, the correlation value becomes high and masks the waveform change that is truly desired to be observed. . In view of this point, the embodiment of the present invention measures the degree of coincidence by subtracting waveforms from each other. r (n) = (1 / T) Σ | {X (ni) −X (nTi)} / V (n) | (2) V (n) = (1 / T) Σ | X (nTi) -A (n) | (3) A (n) = (1 / T) ΣX (nTi) (4) where Σ in equations (2) to (4) is the cumulative sum from i = 0 to T-1. It represents. In the embodiment of the present invention, five detection cycles T are prepared for 20 pps and five for 10 pps detection. Therefore, the calculations of the expressions (2) to (4) are performed each time a low-speed sample is input. Do 10 times.

In the difference judgment, the difference calculation result r of equation (2)
The difference is determined using (n). That is, it is determined whether the smallest one of the 10 difference calculation results r (n) satisfies the condition of the equation (5) (S5Y) or not (S5N). In equation (5), β is selected to be about 0.3. r (n) ≦ β (5) When the elapsed period until Expression (5) is satisfied is observed using an actual dial pulse waveform, it is within 4 periods. Therefore, in the case of the embodiment of the present invention, if the dial numerical value is "5" or more, the leading digit dial pulse waveform can be detected.
Then, provisionally stored as the period T _a period indicating the minimum r selected in step S5 (n) (S6).

In the pulse shape measurement (S7), the pulse shape of the dial pulse waveform is measured. Low speed sample signal 2
In 7a and 7b, the dial pulse waveform changes sharply as compared with the voice waveform, so that it can be used as a waveform feature. In the pulse shape measurement, firstly, a sample point b having the maximum amplitude between the sample time n and the past time n−2T _a is searched for, and secondly, a predetermined sample is past from the first sample point b. Up to (for example, 20pp
In the case of the s period, the minimum or minimum sample point a between b and b-13 is searched for, and thirdly, from the first sample point b to the future by a predetermined number of samples (for example, 20 pps).
In the period, from b to b + 13), the minimum or minimum sampling point c is searched for, and then the left wave height ratio H ₁ = X
(a) / X (b) , stores the right height ratio _{H 2 = X (c) /} X (b), the left waveform width W ₁ = b-a and a right waveform width W ₂ = c-b as pulse shape . However, if the pulse shape measured at this stage is used as a subsequent waveform identification condition, the condition becomes too severe. Therefore, a value obtained by multiplying the wave height ratio by a relaxation coefficient is stored.

Next, the signal processing in the leading digit dial detecting unit 6 until the dial pulse waveform is cut out and recognized will be described. Since the cycle of the dial pulse waveform is known in step S6, the low-speed sample signal 27a is input when the cycle is 20 pps and the low-speed sample signal 27b is input when the cycle is 10 pps (S8, FIG. 7). Although not shown, one sample input in step S8 is F
The sample input in step S1a or S1b is stored in the IFO-type RAM successively.

At this point, the RAM is 20 pps or 1
In order to make it dedicated to 0 pps, in the embodiment of the present invention, the configuration is changed so as to be able to store a waveform for 16 basic cycles, that is, 400 samples. This change can be easily made by changing the setting method of the pointer for managing the RAM, and the limited RAM resources in the DSP can be effectively used.
It should be noted that the storage capacity of the dial / pulse waveform does not need to be 16 basic cycles, and even if cycle fluctuations are taken into consideration,
It is sufficient to have one basic cycle. However, since the dial pulse waveform includes the mute pulse waveform generated when the dial mute operation is turned on / off in the rotary dial telephone or the pulse-shaped voice waveform that exists in the immediate vicinity of the dial pulse waveform, the dial pulse waveform is included. Until the recognition is done, a storage capacity with a margin is required.

In the end determination (S9), the end point of the dial pulse waveform is detected. The end time is the time when a pulse waveform satisfying the pulse shape no longer appears. The pulse shape is determined by comparing it with the value obtained in step S7 centering on a predetermined sample number, for example, the sample time point n-13 of the past 13 samples. If the equations (6) and (7) are satisfied on the left and right with respect to the sampling time point n-13, the pulse waveform is present (S9N). X (n-13-W ₁ ) / X (n-13) ≧ H ₁ (6) X (n-13 + W ₂ ) / X (n-13) ≧ H ₂ (7) performed for the signal, if a predetermined time (e.g. 1.5T _a) continuously not O and there pulse, it is determined that there is no pulse waveform at that time, and stops the input of a low-speed sample signals, predated 1.5T _a past The time point is set as the end of the dial pulse waveform (S9Y).

In the tip detecting operation (S10), the dial
The tip of the pulse waveform is detected. When detecting the tip, R
Return to 400 samples before the beginning of the waveform stored in the AM, that is, the end in the embodiment of the present invention, and perform step S
The pulse shape is measured as in 7. That is, the search for the maximum value is started from the position 400 to 13 samples returned from the end, and if the maximum value is found, the minimum or minimum value is searched for the left and right 13 samples, and the left and right wave height ratios and the left and right waveform widths are the values stored in step S7. And the sample point that first satisfies the condition is temporarily set as the leading pulse waveform. Further measures between pulse shape of 1.5T _a termination side from the head pulse waveform provisional and true head pulse waveform if pulse. This reduces the probability that a voice waveform approaching the dial pulse waveform will be mistaken for the leading pulse waveform. The tip of the dial pulse waveform is the minimum or minimum sample point on the left side of the leading pulse waveform.

In the waveform classification operation (S11), dial
Accurate period detection of pulse waveform and waveform division. Cycle detection is performed with square waveforms and dials stored in RAM.
The product sum s (t, p) with the pulse waveform is calculated by the equation (8), and it is taken as the period of the square waveform when the value becomes maximum.

S (t, p) = (1 / K) Σ {ΣX (jt + i + p) + ΣX (jt + i + p)} (8) Here, the first Σ is j = 0 to K−1. The second Σ is i = 0 to L−1, and the third Σ is i = t−L to t
It represents the cumulative sum up to -1. In equation (8), t
Is a period in the embodiment of the present invention, 23 ≦ t ≦ 27, p is a phase value with 0 at the tip of the dial pulse waveform obtained in step S10, 0 ≦ p <t, and K is the number of samples from the tip to the end. Is divided by the period t and the remainder is discarded, L is the number of samples in the section where the square waveform is set to 1, and in the embodiment of the present invention, 8 is the dial pulse waveform stored in the RAM. Equation (8) is calculated for all combinations by changing t and p according to the order, and t and p when the maximum is obtained are the optimum period T and phase P. Regarding the cycle, the content T _a stored in step S6 is corrected to T. Regarding the phase, the phase P is added to the tip sample point obtained in step S10 to make the tip again. The division of the waveform is equivalent to dividing the waveform into 0s at the tip and a period T.

FIG. 8 shows how the waveform classification processing in step S11 is performed when the value of the equation (8) is maximum. In the figure, (a) is a dial pulse waveform stored in RAM, (b) is a square wave having a period of the dial pulse waveform stored in RAM, and (c).
Indicates that the waveform is divided at a cycle T with a new tip sample point being 0.

In the unnecessary waveform elimination operation (S12), when the unnecessary waveform which is the mute pulse waveform or the pulse voice waveform is near the dial pulse waveform, it is excluded and the dial numerical value is recognized. The range to be excluded is two cycles. That is, even if there is a mute pulse or the like, it is at most one cycle, and it is sufficient to consider two cycles. The exclusion determination of FIG. 8A is independently performed for the leading end and the trailing end. In the method described below, if the first cycle or the second cycle is the dial pulse waveform, the dial pulse waveform is surely the third cycle or the position corresponding to the change point of the break period. The fact that almost the same power as in the 4th cycle is observed is used.

In the exclusion judgment on the tip side, first, 3 from the tip.
The powers P _f3 and P _f4 in the front 1/3 section are measured for the third cycle T ₃ and the fourth cycle T ₄ , and the smaller value is set as the front reference power P _f, and the power P _b3 in the rear 1/3 section, P _b4 is measured and the smaller value is used as the rear reference power P _b . Secondly, for the first period T ₁ and the second period T ₂ , the powers P _f1 and P _f2 in the front 1/3 section are measured, and the power P 1 in the rear 1/3 section is measured.
_b1 and _Pb2 are measured. _Thirdly , P _f1 for the first cycle T ₁
If> γP _f and P _b1 > γP _b , it is regarded as a dial pulse waveform, and the others are excluded. Select γ around 0.4. Fourthly, when the first cycle T ₁ is excluded, the same exclusion determination as that of the first cycle T ₁ is performed for the second cycle T ₂ (see FIG. 9).

Next, in the exclusion judgment of the terminal part, the first and second cycles from the terminal end are excluded as in the case of the tip part, and the front and rear electric powers of the third and fourth cycles from the terminal end are respectively measured and small. One is set as the reference power, and the first cycle and the second cycle from the end are compared to determine whether to exclude.

FIG. 9 shows a divided waveform (FIG. 8C) in the case where the mute pulse is excluded at the tip in the operation of step S12 already described, and the mute pulse to be excluded has one cycle. In eye T ₁ . The front and rear reference powers P _f and P _b are measured from the waveforms of the third cycle T ₃ and the fourth cycle T ₄ , and the powers P _f1 and P _{b1 of the} first cycle T ₁ are measured.
Compared with P _f1 , the front side P _f1 satisfies the condition of P _f1 > γP _f , but the rear side P _b1 does not satisfy the condition of P _b1 > γP _b because P _b1 = 0, so the first cycle T ₁ The mute pulse having the waveform of is excluded (S12). Since the dial numerical value excludes the unnecessary waveform that is not counted as the dial pulse waveform in step S12 from the number of cycles (the number of pulses) before exclusion of unnecessary waveform found in step S11,
It is a value obtained by subtracting the period number 1.

In the next step S13, it is determined whether or not the waveform is abnormal. In other words, when voice is mixed in during dialing, or when the dial pulse waveform fluctuates in time due to telephone line conditions, the waveform stored in RAM is used as the reference waveform to detect subsequent digits. Cannot be used. Also, the recognized dial digit of the first digit itself is not reliable. Therefore, with respect to the dial pulse waveforms that are left without being excluded, abnormal waveform detection is performed as follows. First, the front side of the first cycle 1 /
The power P _{f in the} three sections and the power P _b in the rear 1/3 section are used as reference power. Next, the power P _fx of the front 1/3 section and the power P _bx of the rear 1/3 section after the second cycle are measured to _obtain the reference power P.
Compare with _f (see FIG. 9). That is, four comparisons, P _fx <δ · P _f P _fx > ε · P _f P _bx <δ · P _b P _bx > ε · P _b are performed, and if any one is satisfied, it is regarded as an abnormal waveform. For the coefficients, for example, δ = 0.2 and ε = 5 are selected.

Then, in step S14, the leading digit dial numerical information 28 is output from the leading digit dial detecting section 6 to the PB signal sending section 8. However, when it is determined that the waveform is abnormal, information other than the dial numerical value, for example, a dial signal B that is not normally used (0 to 0 for dial signals) is used.
In addition to the numerical value of 9, there are several signals including B. ) Is output.

In the next step S15, the number of samples per cycle is k, the number of cycles excluded on the leading end side is m, and step S11
A new tip sample point p + m, where n is the tip sample point p obtained in step S and the dial number recognized in step S12
Store the samples from k to the new end sample point p + mk + nk as a dial pulse waveform. Further, a sample for one cycle from the end sample point is stored as a trailing waveform (see FIG. 10). The tail waveform is an isolated waveform (dial 1) in the recognition of the trailing dial pulse waveform.
It is used to increase the recognition accuracy of.

Since the dial pulse waveform and the tail waveform stored in step S15 are used as the reference waveforms in the subsequent digit recognition, they are frequently read from the RAM.
Therefore, the storage position is rewritten so that the leading edge sample is stored at a predetermined RAM address, so that addressing can be easily performed. The storage capacity required in step S15 is 1
Since it is one cycle, in the embodiment of the present invention, 11 × 27 samples are enough, and 4 × 4 samples used for waveform storage in step S8.
About 100 out of 00 samples are left over at this stage. This is used to store the pulse shape or subsequent digit samples in the subsequent operation to effectively use the DSP internal resources.

In step S16, the average amplitude required for comparing the reference waveform stored in step S15 and the subsequent digit waveform is calculated and stored. The average amplitude is obtained in each section obtained by dividing one cycle of the reference waveform into three. For the following description, a waveform divided in units of one cycle T is called a unit waveform,
A waveform obtained by dividing a unit waveform into three is called a divided waveform. The equation for calculating the amplitude average of the front side divided waveform is represented by the equation (9), the center divided waveform is represented by the equation (10), and the rear side divided waveform is represented by the equation (1).
It is shown in 1).

P _f (k) = (1 / L) Σ | X ((k−1) T + i) −V (k) | (9) where Σ is from i = 0 to L−1. It represents the cumulative sum. P _c (k) = (1/8) Σ | X ((k-1) T + i) −V (k) | (10) where Σ represents the cumulative sum from i = L to L + 7. There is. P _b (k) = (1 / M) Σ | X ((k-1) T + i) −V (k) | (11) where Σ is the cumulative sum from i = L + 8 to T−1. It represents. In the equations (9) to (11), k is a serial number of the unit waveform in which the first unit waveform is 1, and the upper limit is a value obtained by adding 1 to the recognized dial number, L = (T-8) / 2 (decimal point and below). (Truncated), M = T−L−8, T is the period, X is the reference waveform, and V (k) is the average value of one period shown in equation (12).
In the embodiment of the present invention, the average section length at the center is set to about 1
This is because it corresponds to / 3 cycle. V (k) = (1 / T) ΣX ((k-1) T + i) (12) Here, Σ represents the cumulative sum from i = 0 to T-1.

FIG. 10 is a waveform diagram showing the pulse shape measuring operation of the leading digit dial detector 6. Step S17
In FIG. 7, the pulse shapes of three parts in the reference waveform (cycles T _{1 to} T ₆ ) are measured and stored.

First, the left waveform width W ₃ , the left wave height ratio H ₃ , the right waveform width W ₄ , and the right wave height centering on the peak (time point b ₁ ) of the front divided waveform of the head unit waveform existing in the cycle T _1. Find the ratio H ₄ . Here the left wave height ratio H ₃ time a ₁ amplitude X of (a ₁₎ of the point b ₁ is amplitude X (b _1), i.e., since a value obtained by dividing the amplitude of the peak, H ₃ = X (a ₁ ) / X (b ₁ ). The right wave height ratio H ₄ also becomes H ₄ = X (c ₁ ) / X (b ₁ ) using the amplitude X (c ₁ ) at the time point c ₁ .

Second, the left waveform width W ₅ and the left wave height ratio H ₅ are centered on the largest peak value (time point b ₆ ) of the front side divided waveforms in each period from the period T _{2 to} the end of the period T _6. , Right waveform width W ₆ and right wave height ratio H ₆ are obtained.

Thirdly, the left waveform width W ₇ and the left wave height ratio H are centered on the largest peak value (time point b ₃ ) in the rear divided waveforms in each period from the period T _{2 to} the end of the period T _{6. 7,} right waveform width W _8, obtain the right height ratio H _8.

Here, H _{5 to} H ₈ are time points a ₆ , b ₆ , and c.
Amplitudes X (a ₆ ), X (b ₆ ), and X (c of ₆ , a ₃ , b ₃ , and c _3
6 ), X (a ₃ ), X (b ₃ ), and X (c ₃ ) are used. H ₅ = X (a ₆ ) / X (b ₆ ) H ₆ ═X (c ₆ ) / X (b _{6 ) H 7 = X (a 3} ) / X (b 3) H 8 = X (c 3) / X to become (b _3).

The reason for having three measurement portions is that the leading pulse waveform (T _{1 in} FIG. 10) is different from the following pulse shape due to the transient response, and the front side divided waveform and the rear side divided waveform have different response characteristics. Therefore, each representative, that is, the maximum portion (T _{6 in} FIG. 10) of the front side divided waveform and the maximum portion (T _{3 in} FIG. 10) of the rear side divided waveform are required. The representative is set to the maximum peak portion because it is easy to select. As described in the processing of step S7, it is too strict for the wave height ratio H to be used as the determination condition, so the wave height ratio H is corrected to a value multiplied by the relaxation coefficient.

In step S18, it is determined whether or not the trailing waveform (FIG. 10) for one cycle following the end of the reference waveform should be compared when recognizing the subsequent digit. When the unnecessary waveform is excluded on the terminal end side of the waveform in step S12 and immediately before the end of the reference waveform, one cycle power of the unit waveform (cycle T in FIG. 10).
_{If the} tail waveform power (FIG. 10) is equal to or more than a predetermined value (for example, −6 dB or more) compared to the power of ₆ ), the tail waveform is regarded as a mute pulse and the tail waveform comparison can be performed in the subsequent digit recognition.

During the signal processing in steps S10 to S18, there is a silent signal period corresponding to the digit between the first dial and the second dial, and the digit gap is 20 pps dial.
There is an interdigit time of 450 ms or more in the case of a pulse waveform and 600 ms or more in the case of a 10 pp dial pulse waveform.
Therefore, in the processing of steps S10 to S18, the input of the low-speed sample signal 27 may be stopped and the processing may be completed within 400 ms.

Next, the processing until the first cycle is detected in the subsequent digit recognition will be described with reference to FIGS. 11 and 12. In step S19, the low speed sample signals 27a, 27
Input one sample of b. The input samples are stored in a FIFO type RAM (not shown). The storage capacity is
The maximum period is 27 samples, and 13 samples for measuring the left and right pulse shapes are added, which is 53 samples. In step S20, the basic waveform and the input waveform are compared. The comparison is performed independently for the three parts of the front divided waveform, the center divided waveform, and the rear divided waveform. The difference calculation for this is expressed by the formula (1
3) to (15).

E _f (k) = (1 / L) Σ [| Y (n-12-T + i) −X ((k-1) T + i) |] / P _f (k) (13) Here, Σ represents the cumulative sum from i = 0 to L-1. E _c (k) = (1/8) Σ [| Y (n-12-T + i) -X ((k-1) T + i) |] / P _c (k) (14) where Σ Represents the cumulative sum from i = L to L + 7. E _b (k) = (1 / M) Σ [| Y (n-12-T + i) -X ((k-1) T + i) |] / P _b (k) (15) where Σ Represents the cumulative sum from i = L + 8 to T-1.

In equations (13) to (15), n represents the current sample time (sample position), L, M, T and X are the same as those in step S16, and Y is input in step S19. Stored waveform, P _f (k), P
_c (k) and P _b (k) is the average amplitude of the divided waveform stored in step S16. k is the same serial number as k in step S16, but the upper limit value is a value obtained by subtracting 1 from the upper limit value. Y
12 in the pointer formula for convenience of pulse shape measurement is for knowing the sample value on the right side from the peak value of the rear side divided waveform. After the comparison operation is completed, the minimum value of E _f (k) is selected, and it is set to the minimum value E _f , also E _c (k).
Is the minimum value E _c , and E _b (k) is the minimum value E _b ,
The difference value r (n) is calculated as the average value of the three as shown in equation (16). _{r (n) = (E f} + E c + E b) / 3 (16)

The reason for performing such a difference calculation will be described. Normally, the dial pulse waveform and the subsequent dial pulse waveform are very similar to each other. Therefore, if the comparison is made in units of one cycle, the subsequent dial pulse waveform can be detected and the voice waveform and the like can be eliminated. However, in rare cases, the tip waveform is different from that of the leading dial pulse waveform, or when the number of waveforms of the leading dial pulse waveform is smaller than the number of subsequent waveforms, for example, when the leading digit is "5" and the trailing digit is "0". Pulse waveforms that do not exist in the reference waveform created from the first digit may be included in the pulse waveforms in the succeeding digits, and the conditions are too strict for comparison on a cycle-by-cycle basis. I can't do it and overlook it. A simple method for solving this is to loosen the determination standard of the difference calculation result, but the defense against the voice waveform becomes loose. Therefore, in the embodiment of the present invention, one cycle T is divided into three parts so that comparison can be made in divided waveform units (see FIG. 9), and a synthetic one-cycle waveform composed of a combination of divided waveforms is to be compared. Thus, by preserving the characteristics of the reference pulse waveform and creating a large number of reference waveforms, it is possible to avoid overlooking the pulse waveform of the subsequent digit.

In step S21, a difference judgment is made. That is, if the difference value r (n) obtained in step S20 is less than or equal to ζ, it is determined that they match and the process proceeds to a new step (S21
Y), otherwise enter the next sample (S21)
N). At this time, ζ is selected to be about 0.5. Step S2
In No. 2, when the voice waveform and the like pass the difference determination, the pulse shape is measured and determined to eliminate the voice waveform and the like. Since the RAM has a waveform storage capacity for two cycles, 13 samples are stored on the left side of the front divided waveform and 13 samples are stored on the right side of the rear divided waveform, so that the pulse shape can be measured (see FIG. 9). Find the peak value of the front side divided waveform (the position is P ₁ (corresponding to b ₁ of FIG. 10)) and the rear side divided waveform (the position is P ₂ (corresponding to b ₃ of FIG. 10)) and center it Then, the pulse shape is compared with the pulse shape stored in step S17.

That is, if the expressions (17) to (20) are satisfied, it is determined that the waveform is a dial pulse waveform. Y (P ₁ -W ₃ ) / Y (P ₁ ) ≧ H ₃ (17) Y (P ₁ -W ₄ ) / Y (P ₁ ) ≧ H ₄ (18) Y (P ₂ -W ₇ ) / Y (P ₂ ) ≧ H ₇ (19) Y (P ₂ -W ₈ ) / Y (P ₂ ) ≧ H ₈ (20) When these equations satisfy the front and rear pulse shapes, dial pulse waveform To consider.

In step S23, the counter is initialized. The counter becomes 0 at the division position of the unit waveform and becomes 1.
The number of samples up to 2T (rounded up to integer, the same applies below) is counted. However, the detection of the leading edge of the succeeding digit is special, and since the waveform for one cycle that already satisfies the difference condition is stored in the RAM, the value of the cycle counter is 1.0T when it is regarded as a dial pulse waveform. And This is a temporary decision, and when the result of the difference calculation subsequently performed thereafter has a better value, it is re-corrected at that time.

In step S24 (FIG. 12), step S
Input one sample as in 19. In step S25, the counter is incremented. In step S26, the same difference calculation as in step S20 is performed. Step S2
7, the difference value r (n-1) in the previous sample is compared with the difference value r (n) in the current sample, and if r (n-1)> r (n), then the minimum difference value R _v , And the value of the counter at that time is stored as the minimum difference position R _p . Step S2
In step 8, if the counter value is 1.2T, step S
The process proceeds to 29 (S28Y), and if less than that, the process returns to the next input (S24) to check the difference value (S28N). The reason why the upper limit of the range in which the difference calculation is performed is set to 1.2T is that the period of the subsequent digit dial has jitter with respect to the period T obtained in the detection of the first digit dial (S11). This is because it is necessary to detect

In step S29, the counter is initialized using the minimum difference position R _p . That is, 1.2T-R _p is used as the counter value. As a result, the boundary position of the unit waveform becomes 0 as the counter value. In step S30, the pulse shape of the unit waveform ending at the minimum difference position R _p is measured to determine the pulse shape value P _s . The measurement is performed similarly to step S22. Next, when both the front side divided waveform and the rear side divided waveform are pulses, P _s = 2, and when one is not a pulse, P _s = 1 and when neither is a pulse, P _s = 0. Since the pulse shape has already been confirmed in step S22 in step S30, P _s = 2 is always satisfied.

FIG. 13 shows steps S27 to S30.
It is shown that the likelihood (likelihood) is determined from the minimum difference value R _v , the minimum difference position R _p, and the pulse shape value P _{s obtained in} . In step S31, this likelihood information is measured. Likelihood information is used to detect the end of the dial pulse waveform and is measured every time a unit waveform is divided. The likelihood information in step S31 is 2 since it is immediately after the detection of the head unit waveform.

FIG. 14 shows a normal dial with a mute pulse rising at time a having a period from time b to time d.
A waveform is shown in the case where it is detected as a pseudo unit waveform from time points a to c because it approaches the pulse waveform. In such a case, steps S24 to S31 (see FIG. 1).
Since it cannot be eliminated under the leading edge detection condition of the succeeding digit described in 2), it is determined in the processing of the second cycle.

15 and 16 show the flow of the operation of the processing of the second cycle. Step S32, S
At 33, the same processing as steps S24 and S25 is performed. In step S34, control is performed so that only sample values are stored until the counter exceeds the lower limit of the time range in which waveforms are compared (S34N) so as not to unnecessarily capture information. The lower limit value is 0.5T, and if it exceeds this value, the process proceeds to step S35. Steps S35 and S36
Then, the same processing as steps S26 and S27 is performed to obtain the minimum difference value R _v and the minimum difference position R _p .

In step S37, comparison with the tail waveform is performed. The comparison calculation is performed using equations (13) to (15), and k used in step S16 is a value indicating the tail waveform, that is, a value obtained by adding 1 to the dial numerical value recognized at the leading digit. After that, the three calculation results are averaged, and if the average value is, for example, 0.3 or less, the comparison result R _t of the tail waveform is set to 1,
If not satisfied, R _t = 0. However, when R _t = 1 it is not reset to 0. The comparison result of the tail waveform is used at the time of detecting the end as interpolation information of a waveform having a small amount of information such as dial "1".

In step S38, as in step S28, if the counter value is 1.2T or less, control is performed so that the previous processing is repeated (S38N), and the counter value is 1.2.
If it is T (S38Y), the process proceeds to step S39.

In step S39 (FIG. 16), mute /
Exclude judgment of pulse etc. That is, when the mute pulse in FIG. 14 is set as the leading end, determination is made by utilizing the fact that the minimum difference position in the second cycle shifts. When the mute pulse is mistakenly recognized as the tip waveform, the time point a becomes the tip sampling point as shown in FIG. The second cycle is point c
Is counted as 0, but a normal dial pulse waveform exists between the time point c and the sampling point at which the counter reaches 1.2T. Therefore, the minimum difference position R _p is a position that delimits the normal dial pulse waveform, that is, the time point d, which is smaller than the value measured from the time point b when the normal tip sample is obtained. The degree of decrease is almost equal between the time points b and c of the pulse waveform width, and is 0.2T to 0.3T. Therefore, in the embodiment of the present invention, the minimum difference position R _p
Is 0.8 or less, it is determined that the detection is erroneous due to the mute pulse. When it is determined that the detection is erroneous and the waveform should be excluded (S39N), the process proceeds to step S40.
If not, it is assumed that a normal waveform is detected (S39
Y), and proceeds to step S41.

In step S40, the counter value is set to 1.2.
Initialize T-R _p, and 1 dial number. In step S41, the counter value is initialized to 1.2T-R _p and the dial numerical value is set to 2. In step S42, the pulse shape value P _s is obtained as in step S30. In step S43, likelihood information is measured according to FIG. Likelihood information is measured and stored for each unit waveform.
If the pulse is excluded, the likelihood information obtained in step S31 is discarded, and the likelihood information obtained in step S43 becomes the first stored value.

FIG. 17 and FIG. 18 show the flow of operations in the subsequent digit dial detecting section 7 from the third cycle onward of the end detection and dial numerical value recognition. The first half of the processing after the third cycle is almost the same as the processing of the second cycle (FIGS. 15 and 16).
Step S44 is step S32, step S45 is step S33, step S46 is step S34, step S47 is step S35, step S48 is step S36, step S49 is step S37, step S50 is step S38, step S51 is step S51.
42 and step S52 perform the same processing as step S43. In step 53 the counter is initialized, the value is set to 1.2 T-R _p. The minimum difference value R _v is 0.5
In the above case, since it is not possible to clearly divide the unit waveform, the unit waveform is divided into 1.0T and the counter is set to 0.2T. In step S54, the dial numerical value is incremented. In step S55, the termination is determined. The end determination is made based on the arrangement of the likelihood information, and if the sum of the latest three pieces of likelihood information shown in FIG. 13 is 1 or less, the end is determined (S5).
5Y).

When there is no mute pulse at the end of the dial pulse waveform, the likelihood information is arranged in the sequence "00".
When it reaches 0 ", it is the end (sum is 0),
If there is a pulse, "100", "010",
It becomes one of "001" (the sum is 1). The reason for determining the end by the arrangement of the three pieces of likelihood information appears when noise appears in the middle of the dial pulse waveform.
This is to allow sequences such as 01 ”,“ 101 ”, and“ 110 ”. Such a determination cannot be made with the sequence of two pieces of likelihood information. If it is not determined to be the end, step S4.
The operation from 4 is repeated (S55N).

If it is determined to be the end, the process proceeds to step S56. In step S56, the dial numerical value is determined. The dial number is the dial number stored in the previous step minus three. In step S57, dial waveform recognition is performed. That is, if it is determined in step S18 (FIG. 7) that the rear waveform comparison is possible, and the rear waveform comparison result R _t is 1, the dial
It is regarded as a pulse waveform, and when R _t = 0, it is not regarded as a dial pulse waveform and the dial numerical value is discarded. Step S
In 18, if the tail waveform comparison is not enabled, it is regarded as a dial pulse waveform. By performing such recognition processing, the probability of erroneously recognizing a voice signal as a dial "1" is reduced. In step S58, the dial numerical value is output to the PB signal transmitter 8 when it is regarded as the dial pulse waveform. Thereafter, an initialization operation such as clearing of the RAM for storing the input waveform to detect the subsequent dial pulse waveform is performed, and then step S19 (see FIG. 1).
Return to 1).

[0100]

According to the present invention, the following effects occur.

First, since the leading digit dial is detected by the difference calculation using a plurality of cycles, the dial pulse waveform can be detected even if the dial numerical value is not known on the detecting side, and it depends on the receiving level. Not possible to detect.

Secondly, in the first digit detection, the pulse shape at the time of detecting the dial pulse waveform is measured to detect the end of the dial pulse waveform, so that it is possible to detect the end excluding the voice waveform.

Thirdly, in the detection of the leading digit, the unit waveform is accurately distinguished by the product-sum operation with the rectangular waveform, and the comparison judgment with the divided waveform power of the portion which is surely the dial pulse waveform is made. Since this is done, it is possible to exclude the interfering signals such as the mute pulse at the leading end and the trailing end and accurately recognize the dial value.

Fourthly, in the subsequent digit detection, since the make rate of the dial pulse is 30%, one cycle of the reference waveform is divided into three, and the difference between the combined waveform and the input waveform is calculated. By making a determination that the determination criterion of the calculation result is not loosened, it is possible to reduce false recognition of the voice waveform and the like.

Fifth, since the end is detected by using the likelihood information in the subsequent digit detection, the dial numerical value can be accurately recognized without erroneously recognizing the end in the middle of the dial pulse waveform.

Sixthly, in the subsequent digit detection, the minimum difference position in the second cycle is checked, and if it is smaller than a predetermined value, the tip position is corrected. Therefore, a pseudo dial pulse waveform including a mute pulse is obtained. Can be excluded.

Seventh, since the correction is made by the minimum difference position of the initial value of the counter which divides the dial waveform in the subsequent digit detection, it is possible to detect the dial pulse waveform including the jitter.

Eighth, since the comparison with the effective tail waveform in the subsequent digit detection is performed to determine whether or not it is the dial waveform, the voice signal is less likely to be erroneously recognized as an isolated dial pulse waveform. .

Ninth, the sampling speed is reduced by the absolute value calculation and the thinning-out processing, and before the dial pulse waveform is detected, the low-speed sampling signal for 20 pps and 10 pp
An independent means for storing the low-speed sample signal for s is prepared,
After the dial pulse waveform is detected, the configuration is changed so that it becomes an integral storage means, so that it can be realized using an inexpensive DSP with a small storage capacity.

Tenthly, since the abnormal waveform judgment is performed after the division of the leading digit dial waveform and the special dial numerical value is notified to the receiving side system, the system side requests the caller to make a call back or the like. Will be possible.

Eleventh, since the received signal is passed through a high-pass filter having a cutoff frequency sufficiently higher than the low cutoff frequency of the telephone line and a linear phase characteristic, the ringing and the first formant of voice are attenuated. The dial pulse waveform can be detected reliably.

Twelfthly, the average value of the past fixed period and the change amount around the average value are compared, and the waveform with a small change amount is dialed.
Since it is not regarded as a pulse waveform, the PB signal and the like can be eliminated.

Thirteenth, since the two-line / four-line conversion section is used to remove the circling voice, the dial pulse can be correctly detected even during the voice guidance. Therefore, the effect of the present invention is extremely large.

[Brief description of drawings]

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

FIG. 2 is a circuit configuration diagram showing a detailed configuration of a preprocessing unit which is a component of FIG.

FIG. 3 is a waveform diagram illustrating the operation of a high-pass filter and an absolute value calculation unit, which are components of FIG.

FIG. 4 is a circuit configuration diagram showing an internal configuration of a voice removing unit which is a component of FIG.

5 is a waveform diagram of each part of the circuit configuration of FIG.

FIG. 6 is a flowchart illustrating an operation flow of a leading digit dial detection unit which is a component of FIG. 1.

FIG. 7 is a flowchart illustrating an operation flow of the leading digit dial detection unit which is a component of FIG. 1 together with FIG. 6;

FIG. 8 is a waveform diagram for explaining a waveform division operation of the leading digit dial detection unit.

FIG. 9 is a waveform diagram illustrating an operation of excluding a mute pulse on the leading end side of the leading digit dial detection unit.

FIG. 10 is a waveform diagram for explaining a pulse shape measuring operation of the leading digit dial detection unit.

11 is a flowchart showing a flow of a dial waveform detecting operation in a succeeding digit dial detecting unit which is a component of FIG.

FIG. 12 is a flowchart showing the flow of dial waveform detection operation in the subsequent digit dial detection unit that is a component of FIG. 1 together with FIG. 11.

FIG. 13 is a likelihood diagram for explaining a method of determining likelihood information.

FIG. 14 is a waveform diagram illustrating a mute / pulse exclusion operation of a succeeding digit dial detection unit.

FIG. 15 is a flowchart showing a flow of a mute / pulse exclusion operation in a succeeding digit dial detection unit.

16 is a flowchart showing the flow of a mute / pulse exclusion operation in the subsequent digit dial detection unit together with FIG. 15.

FIG. 17 is a flowchart showing a flow of a dial numerical value recognition operation in a succeeding digit dial detection unit.

FIG. 18 is a flowchart showing the flow of dial numerical value recognition operation in the succeeding digit dial detection unit together with FIG. 17;

[Explanation of symbols]

 1 Telephone line interface 2 AD converter 3 DA converter 4 Echo canceller 5 Pre-processing unit 6 First digit dial detection unit 7 Subsequent digit dial detection unit 8 PB signal transmission unit 9 Switch 10 DA converter 11 AD converter 12 System line Interface 21 Telephone line 22, 24, 26, 31 Received signal 23, 25, 32 Transmitted signal 27, 27a, 27b Low speed sample signal 28 First digit dial numerical information 29 Subsequent digit dial numerical information 30 Dial signal 33 System signal line 34 DC loop Confinement information 41 High-pass filter (HPF) 42 Absolute value calculator 43a, 43b Low-pass filter (LPF) 44a, 44b Decimation processor 45a, 45b Speech remover 46, 46a, 46b Low-speed sample signal 51 High-pass Pass filter (HPF) 52 DP average Force adding 53 limiter 54 and 55 signals

Claims (14)

[Claims] 1. A reception signal (26) including a dial signal.
And an absolute value calculation unit (42) for shifting the occupied frequency band to the low frequency side including the DC component by obtaining the absolute value of Pre-processing means (5) including a voice removing unit (45) for outputting a low speed signal (27) from which the signal has been removed, and a plurality of different periods for the components of the dial signal included in the low speed signal. A leading digit dial detecting means (6) for performing a difference operation to detect the waveform of the component of the dial signal, storing the waveform information, and recognizing the dial numerical value of the leading digit, and the stored waveform information and waveform. A dial detecting device including a succeeding digit dial detecting means (7) for comparing and calculating a difference, detecting a waveform of a component of a subsequent dial signal, and recognizing a dial numerical value of a succeeding digit. 2. The dial detecting device according to claim 1, wherein the leading digit dial detecting means recognizes that the end of the waveform is detected when the waveform is not detected within a predetermined period from the time when the waveform information is detected (S9). 3. The leading digit dial detecting means calculates the optimum value of the cycle and the phase by performing a sum-of-products operation of the waveform based on the stored waveform information and a square wave representing the cycle of this waveform, and divides it into unit waveforms. (S11), the power (P _f , P _b ) of the front divided waveform of the front side portion and the rear side divided waveform of the rear side portion in this section is determined to determine whether the mute pulse is excluded (S12). The dial detection device according to claim 1. 4. The leading digit dial detecting means detects waveforms of a plurality of dial signal components having different dial speeds and outputs each waveform information until the dial waveform detection time at which the waveform of the dial signal component is detected. After the dial waveform is detected, the waveform information detected from the components of the plurality of dial signals having different dial speeds is not individually divided but is integrally stored (see FIG. S15) The dial detection device according to claim 1. 5. The succeeding digit dial detecting means, when performing the difference calculation by comparing the waveforms, divides a waveform divided into three parts into a front side, a center, and a rear side of a cycle of the waveform represented by the stored waveform information. The dial detection device according to claim 1, wherein the waveform comparison is performed as a unit (S21). 6. The subsequent digit dial detecting means divides the waveform cycle represented by the stored waveform information into three parts, ie, a front side, a center, and a rear side when the difference is calculated by performing the difference calculation. The counter value for counting the number of sections for partitioning the waveform is compared in the unit of the waveforms, and the counter value measured in a range selected to be wider than the jitter of the period of the dial pulse signal. 2. The dial detection device according to claim 1, wherein the difference value obtained by the waveform comparison and the difference calculation is initialized using the value of the minimum difference position (R _p ) which is the position showing the minimum value (S29). 7. A minimum difference position where the subsequent digit dial detection means is a position at which the difference value obtained by the waveform comparison and the difference calculation is the minimum value when the waveform comparison is performed and the difference calculation is performed. When the second minimum differential position is detected within a predetermined period after the first minimum differential position is detected, the pulse of the first minimum differential position is recognized as a mute pulse and excluded, and the second minimum differential position is detected. The dial detection device according to claim 1, wherein is identified as a new tip sample position (S36). 8. The subsequent digit dial detecting means divides the waveform cycle represented by the stored waveform information into three parts, ie, the front side, the center, and the rear side, when the difference is calculated by the waveform comparison. The waveforms are compared with each other as a unit, and the minimum difference value (R _v ) showing the minimum difference value obtained by the difference calculation and the minimum difference position (R _p ) showing the temporal position of this minimum difference value. And the pulse shape value (P _s ) representing the waveform of the detected subsequent dial signal component, the likelihood information indicating the existence of the pulse waveform is measured (S31). When it shows a predetermined value, it is determined that the pulse waveform is terminated (S31, S4).
3, S52) The dial detection device according to claim 1. 9. The subsequent digit dial detecting means compares the stored waveform information with the stored waveform information to perform a difference operation to detect a waveform of a component of a subsequent dial signal. 2. The dial detecting device according to claim 1, wherein it is determined whether or not it is a dial pulse waveform by comparing the waveform with the tail waveform of the two (S18, S37). 10. When the pre-processing means applies a reception signal including the dial pulse signal to the absolute value calculation unit, the reception signal applied to the absolute value calculation unit is a low frequency band of a telephone line. 2. The dial detection device according to claim 1, comprising a high-pass filter (41) having a substantially linear phase characteristic that cuts off at a frequency sufficiently higher than the cut-off frequency. 11. The dial detection device according to claim 1, wherein the pre-processing means includes a thinning processing unit (44a, 44b) for thinning and sampling the output of the absolute value computing unit to obtain the low speed signal. . 12. The leading digit dial detecting means operates so as not to be recognized as a dial signal when a change amount of the received signal with respect to an average value of the received signals in a predetermined past period is a predetermined amount or less (S5). S21) Claim 1
Dial detection device. 13. When the leading digit dial detecting means cannot detect the waveform of the component of the dial signal in the low speed signal and detects an abnormal waveform, the fact that no dial signal has been detected is notified. The dial detection device according to claim 1, wherein the notification information other than the dial numerical value is output (S13, S14). 14. The dial removing unit (45) removes a component based on a voice signal transferred to a low frequency side from a signal in which the occupied frequency band is transferred to a low frequency side including a direct current component, and the dial. A high pass filter (51) for obtaining a signal component, and a DP average power adder (52) for adding the average power of the dial signal to the component of the dial signal to obtain the low speed signal (27). The dial detection device according to claim 1, further comprising: