JPH012462A - How to distinguish between audio and control signals - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] [Industrial application field]

The present invention relates to a method for identifying a voice signal and a control signal, which makes it possible to distinguish between a "human voice" consisting of a unique frequency spectrum and a "control signal" consisting of a single sine wave. In particular, within the voice frequency band (30011z to 3.4Kllz) such as telephone lines, "human voice" and other connected VCs such as facsimile etc.
The “control signal” of the device (in the case of facsimile, it is generally C
By making it possible to identify the 110011z automatic response signal called an NG signal (hereinafter referred to as a CNG signal), it is possible to make effective use of telephone lines and the like.

[Conventional technology]

An additional device that has one office line connection terminal connected to the telephone line and multiple extension terminal connection terminals connected to telephones and other connected devices (e.g. facsimiles) in order to make effective use of telephone lines. A telephone is used to switch and connect telephones and other connected devices to a single telephone line. In such cases, the additional device identifies the signal from the calling party and switches the telephone line to either the telephone or other connected equipment.
control.

[Problems to be solved by the invention]

However, if the devices connected to the extension terminal connection terminal are telephones, facsimile machines, etc.,
;1. Frequency components of "human voice" from a telephone,
The "control signal" of another connected device may have the same frequency component within the same audio frequency band. For example, “human voice” is 300Hz to 3.4K fiz
The frequency component of the CNG signal is in the audio frequency band, whereas the CNG signal has a frequency of 1100 Hz. In such a case, “
Since "human voice" and "control signal" have the same frequency components, it has been extremely difficult to accurately identify them in an additional device. The object of the present invention is to provide an identification method that can distinguish between "human voice" from a telephone and "control signals" from other connected devices.

[Means to solve the problem]

Detection of a human voice signal involves detecting at least two frequency components F+nP and F− that have a difference of n times the pinch frequency P (where n is an integer) from a formant frequency component F that is equal to the frequency of the control signal. Detection is performed under the AND condition of nP, and the control signal is detected as a frequency component equal to the frequency component of the formant.

[For use]

The control signal is a single sine wave, for example the CNG signal is 11
Although it is composed of a single sine wave of 0011z, human speech is composed of a unique spectrum, and the spectrum is mainly composed of three resonant parts, that is, formants. The pitch frequency P of this formant corresponds to the fundamental frequency of voice, which is approximately 125 Hz for men and approximately 250 Hz for women.
It is. Therefore, by detecting the audio signal under the AND condition of at least two frequency components F+nP and F-nP, which have a difference of n times the pitch frequency P with respect to the formant frequency component F, it is possible to identify it as a control signal. can.

【Example】

Vowels are usually produced by the vocal cords being the sound source, and the shape of the oral cavity giving each vowel a certain shape (this is called articulation). Corresponding to the articulation form of this vowel, specific frequency components are emphasized, such as a, i, tsu, ku. This produces a frequency spectrum that characterizes the five vowels of o. The dominant frequency components that characterize these vowels are called formants, and the formants of the five vowels in Japanese are shown in Figure 2.As is clear from Figure 2, human speech mainly consists of three vowels. It consists of two resonant parts, or formants. In addition, the frequency of voice vibration corresponds to the fundamental frequency of voice, that is, the formant pinch frequency, and the average value and standard deviation of the fundamental frequency for men are approximately 125Hz and 20.5)1z, respectively, and for women it is approximately 250Hz. and approximately 41.51 (z). In this way, human voice is composed of a unique frequency spectrum, but if this human voice and a control signal composed of a single sine wave are combined in a frequency spectrum, The representation is as shown in Figure 3. In Figure 3, (a) is human voice;
(bl indicates the control signal, and the control signal is 11
A CNG signal of 00Hz is shown. As is clear from FIG. 3, since the control signal is composed of a single sine wave, unlike human speech, the control signal has a difference of at least n times the pinch frequency P with respect to the formant frequency component F, unlike human speech. Since the two frequency components F+nP and F-nP do not exist, even if the control signal and the human voice are in the same audio frequency band, the human voice is detected using the AND condition of the frequency components F+nP and F-nP. By doing so, it is possible to distinguish between control signals and human voices. FIG. 1 is a block diagram showing the configuration of an identification circuit based on the above-mentioned principle and which is used in the identification method according to the present invention. 1st
In the figure, BPF1, BPF2, BPF3 are band pass filters, AMPI, AMP2, AMP3, . AMP4 is an amplifier, ADl, AD2, AD3 are waveform shaping circuits, AN
DI and AND2 represent AND circuits, and L represents a logic circuit. The filter BPF 1 is configured to pass a formant frequency component F that is equal to the frequency of the control signal, and the filter BPF 2 is configured to pass a frequency component F+P that is higher than this frequency component F by a pitch frequency P. The filter BPF3 is configured to pass the frequency component FP lower than the frequency component F by the pitch frequency P. Therefore, when assuming a facsimile CNG signal as a "control signal", the filter BPF
The passband of 1 is 1100Hz, and the filter BPF2
The passband of is the male and female pinch frequency (fundamental frequency)
Considering (1100 & +125Hz) = (1100
Hz+250Hz) =1225Hz~1350Hz
Similarly, the passband of filter BPF3 is (110
0Hz -250Hz) ~(1100)1z-125
11z) = 85011z to 97511z. Of course, these passbands will change by that amount if the standard deviation of the fundamental frequency of the voice is taken into account. With this configuration, when a "human voice" signal is applied to the input terminal of amplifier AMP 1, this signal includes frequency components F, F+P, and F-P, so each filter BPF1, BPF2. Outputs are generated from both BPF3, and waveform shaping circuits ADI and AD2
．． AD3 and amplifier AMP2. AMP3. AND circuit AND1. Input is added to AND2. Since the two inputs of the AND circuit ANDI are both 11", its output is "l". Also, the AND circuit AND2
Since there is only one input, which is 61, its output is 61.
''.On the other hand, when a CNG signal is applied as a control signal to the input terminal of amplifier AMP 1, this signal contains only frequency component F, so filter B
Output is generated from PF 1, but filters BPF 2 and B
No output is generated from PF3. Therefore, AND circuit A
Since the two inputs of NDI are both 0'', its output is 0'', but the AND circuit AND2 has only one input, which is ``Bi'', so its output is 1''. In this way, in the circuit configured as shown in FIG.
Since the outputs are different, it is possible to distinguish between "human voice" and "control signal" by configuring a logic circuit according to the purpose of use. Figure 4 shows how to identify the CNG signal and voice of a facsimile.
This figure shows the configuration of a logic circuit that outputs a signal only when an NG signal is input. In Figure 4,
INV is an inversion circuit, OR is an OR circuit, S is an OR circuit
This figure shows an analog switch that turns ON when the output of S is "0." When the logic circuit is configured in this way, the operation of the analog switch S is as shown in the table below. In the described logic circuit, the switch S is turned ON only when the CNG signal is input.
Therefore, the output of this switch S can be used to connect the telephone line to the facsimile. Of course, by changing the configuration of the logic circuit, it is possible to turn on the switch S only when a human voice signal is input, and other configurations can also be implemented as appropriate.

【Effect of the invention】

According to the present invention, a human voice is divided into two frequency components F+nP having a difference of n times the pitch frequency P from a formant frequency component F that is equal to the frequency of the control signal.
, F-nP, it is possible to distinguish the human voice from a control signal that does not have these two frequency components F+nP and F-nP relatively easily and with high reliability. It is possible.

[Brief explanation of drawings]

Fig. 1 is a block diagram of an identification circuit to which the identification method according to the present invention is applied, Fig. 2 is an explanatory diagram of the formants of the five Japanese vowels, Fig. 3 is an explanatory diagram of the frequency spectra of human speech and control signals, FIG. 4 shows a circuit diagram showing an embodiment of the logic circuit. AMPI~AMP 4----Amplifier, BPF 1
~BPF3...-Band pass filter, ADI-AD
3 - Waveform shaping circuit, ANDI, AND2 - AND circuit, L - logic circuit.
,, +7-1 bell (6th) 3rd episode

Claims (1)

[Claims] 1) In systems that identify human voice signals and control signals transmitted via telephone lines, etc. within the same voice frequency band, the voice signal is detected using a formant frequency component F that is equal to the frequency of the control signal. At least two pitch frequencies having a difference of n times the pitch frequency P (where n is an integer)
A method for identifying an audio signal and a control signal, characterized in that detection is performed under an AND condition of two frequency components F+nP and F-nP, and the control signal is detected using a formant frequency component F.