JP2759122B2 - PB band attenuation type voice guidance processing method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a processing method for producing voice guidance used for guiding services using a telephone network such as a message dial.

2. Description of the Related Art In order to use a service of a network (telephone network) such as message dialing which is currently used, a user usually turns a predetermined telephone number for the service, and then turns to the service side. He listens to the voice guidance sent from (Center) and presses the push button according to the instruction to receive the desired service. For example, the message dial has a sequence as shown in FIG.

In such a service, when the user becomes accustomed to the service, the user pushes the push button without listening to the voice guidance to the end. In order to cope with this, it is necessary to operate the push button signal detection device (hereinafter abbreviated as PBR) before the end of the reproduction of the voice guidance.

However, since the current telephone network always has an echo back, if the PBR is operated before the end of the voice guidance, the voice guidance is input to the PBR by the echo back. This is shown in FIG. Voice has a wide frequency band and includes frequencies used by PB signals. Therefore, there is a possibility that the PBR that is operating during the reproduction of the voice guidance with a certain probability detects a PB signal frequency component in the voice guidance as a PB signal, causing a malfunction.

 In order to avoid this, the following method has conventionally been adopted.

That is, this is a method of improving the discrimination performance between a sound and a PB signal by incorporating a circuit for improving the discriminability between the sound and the PB signal into the PBR. However, such PBRs are complex and can easily be expensive. In addition, there is a possibility that the reverse malfunction, that is, an operation in which the PB is regarded as noise even when the user pushes the PB and the PB signal is not detected, may occur. In addition, a large amount of conventional inexpensive PBR is connected to the telephone network. It is practically impossible to replace them.

It is also conceivable to process the voice guidance so that PBR is not detected as a PB signal in the voice guidance.

That is, the voice guidance is processed to attenuate the frequency component of the PB signal so that the PBR does not malfunction depending on the processed voice guidance. This method will be described in detail with reference to FIG.

The guidance 1 of the audio to be processed is digitized and supplied to the Fourier transform unit 2, and is usually subjected to FET (fast Fourier transform) to be converted from the time domain to the frequency domain.
The power spectrum 3 of the voice guidance converted into the frequency domain is passed through a passing filter 4 having a comb-shaped Hanning function characteristic, and the frequency components used by the PB are attenuated in a Hanning function, and other frequency components are attenuated. The power spectrum 5 of the voice guidance thus processed is returned from the frequency domain to the time axis domain by the inverse Fourier transform unit 6 and output as the processed voice guidance 7.

The attenuation characteristic of the comb-shaped characteristic passing filter 4 is as follows.

The PB signal is one of the lower frequencies 697,770,852,941Hz and the upper frequency 1209,1336,1447,1633Hz
4 × 4 = 16 types can be designated by combining with one frequency. And each frequency band is 1.5
%. When each band is attenuated by the Hanning function, the voice guidance is not detected by the PBR because the frequency used by the PB is attenuated. Also, in this processing, distortion is hardly generated in the voice. This Hanning function is expressed mathematically as follows.

f <f0 lwo: I (f) = 0.0 f0 low <f <f0 high: I (f) = (1.0−cos (2.0π (f−f0 low) / f0 rang)
e)) / 2.0 f0 high <f: I (f) = 0.0 where I (f) is an attenuation function, f is a frequency, f0 is a center frequency range is a value that determines an attenuation bandwidth, and is, for example, 0.06; The attenuation bandwidth is range × f0 f0 low = f0 × (1-range) f0 high = f0 × (1 + range) f0 range = f0 high−f0 low f0 = 697,770,852,941,1209,1336,1447,1633Hz As can be seen from this function However, this attenuation curve does not completely attenuate the frequency used by the PB. For example, ra
When ngh = 0.06 is selected, the level in the frequency band used by PB is as follows.

That is, about 8% remains in the frequency band used by the PB without being cut off. However, it is considered that the residual audio power of 8% is not detected as a PB signal by PBR, but may be detected as a PB signal under certain conditions.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a processing method for generating voice guidance in which there is no possibility of being erroneously detected as PB and distortion does not occur.

[Means for Solving the Problems] The present invention completely attenuates at least one of the frequencies used by the PB from the frequency band in which the voice guidance has, and the completely attenuated outer frequency is trigonometric. To attenuate.

"Embodiment" Embodiment 1 An embodiment of the present invention is shown in FIG. 1, and portions corresponding to those in FIG. 5 are denoted by the same reference numerals. The voice guidance 1 processed in the same manner as in FIG. 5 is Fourier-transformed and converted from the time domain to the frequency domain, and the frequency-converted voice power spectrum 3 is completely attenuated in the present invention. Then, the outside of the completely attenuated frequency is passed through a passage filter 11 having a characteristic of attenuating trigonometrically with a Hanning function. Details regarding the filter 11 will be described below. The output of the filter 11 is inverse Fourier transformed to return from the frequency domain to the time domain.

The guidance of the audio processed by the above operation is PB
Will be a sound that does not include any frequency used by. Therefore, even if this voice guidance is input to the PBR, the PBR does not malfunction. Also, since both sides of the complete attenuation band are attenuated by the Hanning function, almost no distortion occurs in voice.

Next, the attenuation characteristics of the filter 11 will be described. As described above, the PB signal is one of the lower frequencies 697,770,852,941 Hz and the upper frequencies 1209,1336,1447,1633H.
This is a combination of z with one frequency, and each frequency band is 1.5%. The filter 11 completely attenuates each of these frequency bands, and attenuates the boundary with the frequency that is not attenuated by the Hanning function. To be precise, only one side of the Hanning function is used. This is expressed by the following equation, and the attenuation characteristic for one frequency is shown in FIG.

f <f0 low low: I (f) = 0.0 f0 low low <f <f0 low center: I (f) = (1.0−cos (2.0π (f−f0 low low) / f0 l)
ow range)) / 2.0 f0 low center <f <f0 high center: I (f) = 1.0 f0 high center <f <f0 high high: I (f) = (1.0−cos (2.0π (f−f0 high low ) / F0
high range)) / 2.0 f0 high high <f: I (f) = 0.0 where f0 low low = f0 × (1−0.015−range) f0 low center = f0 × (1−0.015) f0 low high = f0 × (1−0.015 + range) f0 low range = f0 low high−f0 low low f0 high low = f0 × (1 + 0.015−range) f0 high center = f0 × (1 + 0.015) f0 high high = f0 × (1 + 0. 015 + range) f0 high range = f0 high high−f0 high low f0 = 697,770,852,941,1209,1336,1447,1633Hz range = 0.03 range is a trigonometric function at the frequency not used from the boundary between the frequency used by PB and the frequency not used. The width of the band to be attenuated.

f is the frequency I (f) is the attenuation curve As can be seen from this function and FIG. 2, this attenuation curve completely attenuates the frequency used by the PB. Also,
Since the boundary between the frequency used by the PB and the frequency not used is connected by a Hanning function and is gentle, the side lobe at the boundary between the attenuation region and the non-attenuation region becomes extremely small. Therefore, a speech with almost no distortion is obtained.

Second Embodiment Although the voice guidance processing method shown in the first embodiment causes almost no deterioration in sound quality, it is possible to further reduce the deterioration in sound quality by changing the range value in the first embodiment. In other words, when this value is reduced, the deterioration of sound quality is reduced, but when it is too small, the attenuation characteristic curve between the attenuation region and the non-attenuation region becomes steep, that is, a curve close to a right angle to the frequency axis, The frequency side rope becomes noticeable and beats are heard. vice versa
If the range is increased, the sound will be muffled without high frequencies.

Embodiment 3 To further enhance the voice quality of the voice guidance, a Blackman function may be used instead of the Hanning function as the characteristic of the passing filter 11 in FIG. It is known that the Blackman function has fewer side ropes than the Hanning function. Therefore, it is presumed that there is less quality deterioration. Blackman's function looks like this:

f <f0 low low: I (f) = 0.0 f0 low low <f <f0 low center: I (f) = (0.423−0.498cos (2.0π (f−f0 low lo
w) / f0 low range)) + 0.0792cos (2.0π2.0 (f-f0
low low) / f0 low range)) f0 low center <f <f0 high center: I (f) = 1.0 f0 high center <f <f0 high high I (f) = 0.423−0.498cos (2.0π (f−f0 high lo
w) / f0 high range)) −0.0792cos (2.0π2.0 (f−f0 high low) / f0 high r
ange)) f0 high high <f: I (f) = 0.0 Fourth Embodiment In the first embodiment, all the frequency components of the PB signal are attenuated, but only four of the low-band frequencies in the PB signal, or Only four of the high frequencies may be attenuated. Since the PB signal is composed of a combination of a low frequency and a high frequency, the PBR is not detected as PB unless one of them is detected. By doing so, the power of the attenuated frequency in the voice guidance is reduced, and thus the deterioration of the voice quality of the processed voice guidance is reduced.

Fifth Embodiment To further reduce the deterioration of voice quality, the following may be performed. As a practical matter, even if voice guidance is input to the PBR, not all PB frequencies are detected by the PBR. Therefore, when voice guidance is actually input to the PBR and, for example, "*" is detected in the PBR, one or both of 941 Hz and 1336 Hz of the voice guidance frequency may be attenuated.

Although the voice quality is degraded by the above operations, it is possible to create voice guidance that does not cause any problem in practical use.

[Effects of the Invention] As described above, according to the present invention, voice guidance that does not include the frequency used by the PB and has no distortion can be created. As a result, even if this voice guidance is input to the PBR, no operation occurs. Therefore, PB can be used even when the voice guidance is being reproduced. As an application field, in a network service that uses voice guidance and PB, such as message dialing, it can be used for voice guidance processing when PB and voice guidance need to coexist.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a diagram showing an example of an attenuation characteristic for one frequency in a PB signal of the passing filter 11, and FIG. 3 is a message currently used. FIG. 4 is a diagram showing a sequence of dialing, FIG. 4 is a diagram simply showing the cause of inputting voice guidance to the PBR, FIG.
The figure is a block diagram showing a proposed voice guidance processing method.

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁶ Identification code FI H04M 3/50 H04M 3/50 Z

Claims (1)

(57) [Claims] 1. A system for detecting a push button signal input by a user during voice guidance reproduction via a telephone line, wherein at least one of the frequency bands used by the push button signal is completely attenuated from the voice guidance, and ,
A PB characterized by attenuating frequencies outside the frequency band of the completely attenuated push button signal by a trigonometric function.
Band attenuation type voice guidance processing method.