JPH03175497A - Interrupting device for voice synthesis - Google Patents

Abstract

PURPOSE:To interrupt voice synthesis automatically and restart the voice synthesis later by comparing a difference when a difference due to environmental conditions is minimized and detecting a sudden noise. CONSTITUTION:If a synthesized voice can not be heard suddenly owing to an ambient noise and the voice synthesis is interrupted temporarily, either or both of the output voice of a voice synthesizer 1 and a synthesized voice collected through a microphone are passed through an acyclic filter including characteristics of a speaker, a microphone, etc., and corrected and the differences of the both when differences due to environmental conditions are minimized are compared with each other to detect a sudden noise. When the noise level is larger than a specific threshold value, a voice synthesis interrup tion signal (1) is generated. Consequently, the voice synthesis is interrupted automatically and then restarted afterward.

Description

[Detailed Description of the Invention] [Summary] Automatically interrupting speech synthesis regarding a method for temporarily interrupting speech synthesis when the synthesized speech suddenly cannot be opened due to ambient noise while listening to synthesized speech. For example, for the purpose of improving the operability during operation of a parts collection system, for example, in a speech synthesis device such as an accumulation synthesis (analysis synthesis) type or a rule synthesis type, the output from the speech synthesis device Either or both of the voice and the synthesized voice recorded through the microphone are applied with an acyclic (FIR type) filter that includes the environmental characteristics of the speaker, microphone, etc., to correct the signal and calculate the differences due to the environmental conditions between the two. A sudden noise is detected by comparing the difference when the noise level is minimized, and if the noise level is higher than a certain threshold, an interrupt signal is generated to the speech synthesizer, thereby starting the speech synthesis. automatically interrupt,
Thereafter, the configuration is made to restart speech synthesis.

[Industrial application field]

The present invention relates to a method for interrupting speech synthesis, which temporarily interrupts speech synthesis and then resumes synthesis when the synthesized speech cannot be opened due to sudden noise while listening to the synthesized speech.

Conventionally, parts collection systems using synthesized voices have been known, but when external noise such as a telephone ring occurs,
The synthesized voice may become inaudible, and parts may not be collected.

In order to interrupt speech synthesis in this case, there is an interrupting device that can automatically interrupt speech synthesis without manual intervention.
For example, even if the parts collector's hands are occupied, voice synthesis can be interrupted, improving the operability of the parts collection system and the like.

[Prior art and problems to be solved by the invention] FIG. 9 is a diagram illustrating a conventional speech synthesis interruption method.

Conventionally, in a parts collection system, etc., when listening to synthesized speech from a speech synthesizer l, if it becomes temporarily inaudible due to ambient noise, etc., the interrupt switch 3 is manually operated to interrupt the speech synthesis. was.

Therefore, when working with hands, it is necessary to stop the work and operate an interrupt switch, which poses a problem of poor operability.

In view of the above conventional drawbacks, the present invention provides a device that automatically interrupts speech synthesis using a conventional interrupt switch, detects noise that would make it difficult for the user to hear the synthesized speech, and automatically interrupts speech synthesis. The purpose is to

[Means to solve the problem]

The above problem is solved by a speech synthesis interrupting device configured as follows.

(]) A microphone that records the synthesized voice output from the speaker of the voice synthesizer, an analog/digital converter (A/D) that converts the output of the microphone into a digital signal, and a microphone that records the synthesized voice output from the speaker of the voice synthesizer. An analog/digital converter (A/D) that converts an input analog signal into a digital signal, and a signal corrector that delays and amplifies (or attenuates) the output of the analog/digital converter (A/D) for a certain period of time. , a difference signal for calculating the short-time power of the difference between the output of the signal correction section and the output obtained by digitally converting the input analog signal from the voice synthesizer or microphone by the analog/digital conversion section (A/D). The apparatus is configured to include a power calculation section and an interruption signal generation section that generates a speech synthesis interruption signal when the output of the difference signal power calculation section exceeds a threshold value.

(2) The speech synthesis interrupting device is configured to use an acyclic (FIR type) filter as the signal correction section.

(3) In the above speech synthesis interrupting device, the analog/digital converter (A/D) from the speech synthesizer is not provided, and the digital/analog converter (D) in the speech synthesizer is not provided.
/A) is configured so as to directly pass the digital input to the signal correction section.

(4) In the above speech synthesis interrupting device, each analog/
From the output of the digital converter (A/D), the above acyclic C
The present invention is configured to include a coefficient calculation unit that calculates optimal coefficients of the FTR type filter.

(5) In the speech synthesis interrupting device, a signal correction section that delays and amplifies (attenuates) the output of the microphone for a certain period of time is used instead of the signal correction section that corrects the output from the speech synthesis device.

[Effect]

That is, according to the present invention, in order to automatically detect noise during speech synthesis, for example, the quality of the synthesized speech itself.

Compare the quality of synthesized speech that includes characteristics (including noise) of various surrounding environments such as speakers and microphones.

Therefore, a filter is applied to the synthesized voice that includes the characteristics of the speaker, microphone, etc. This is performed by the signal correction section, and its coefficients are determined by the coefficient calculation section. For example, the following method can be used to determine the coefficient.

As an example of the signal correction section, an acyclic type (Finite 1
There is a filter of npulse Re5ponse (F r R type).

This configuration is that shown in FIG.

Here, the input Xfi to the signal correction section is the data of the synthesized sound itself, and Y, , is the synthesized speech extracted from the microphone.

Noise is detected by comparing this Y,1 with the output Y, . . . of the signal correction section. Coefficient α of this signal correction section.

α1. ..., α9 is calculated by the coefficient calculating section.

In order to efficiently extract true noise, it is necessary to adjust the quality of the synthesized speech to be compared, depending on the environment at the start of speech synthesis, including delay between the speaker and microphone and stationary noise. This is because it is necessary.

When correcting the input signal X7, the characteristics of the speaker and microphone and stationary noise must be taken into consideration. In the coefficient determination phase, the corrected Y, l and the synthesized voice Y from the microphone,
Coefficients are determined to minimize the difference in l. That is, coefficients are determined to minimize noise other than true noise, such as noise dominated by environmental conditions.

From Figure 8, Y e□ =X Hαo + X n-+α, -, +
X1l-, α, 2+ --+ Xl1-, αn-p. The error ε at this time is.

ε= Σ (Y, l −Y, l・)2 ・ Σ (Y, −X, 1α. −X, 1−1α71・ ・
-X R-pαa-p )z . Each coefficient (α0
~α9) is determined. Therefore, let ε be each coefficient (α. ~ α9
) and set it to zero, each coefficient (α.~
αp) is determined. That is, regarding each coefficient,・0(p
), the simultaneous equations consisting of p+1 equations (0)2 to (p) are solved, and each coefficient (α. to α9) is determined.

Using the coefficients determined in this way, for example, it is possible to bring (correct) the digital conversion signal of the analog input signal from the speech synthesizer to the digital conversion output signal of the synthesized voice from the microphone, and to make it true. This makes it easier to extract noise.

Then, the correction signal and the digitally converted signal from the microphone are compared in power for a short period of time to determine the true noise that exceeds a specific threshold value and is used as an interrupt signal for the speech synthesizer.

Therefore, in the present invention, while listening to synthesized speech, noise that occurs in the surrounding environment and is different from the synthesized speech is automatically extracted, and the extracted signal is used as an interruption signal. This has the effect of automatically interrupting speech synthesis.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a diagram showing one embodiment of the present invention, FIG. 2 is an operation flow diagram of the present invention, and FIGS. 3 to 7 are diagrams showing other embodiments of the present invention. For example, if the digital conversion signal of the output voice from the voice synthesizer 1 is
．． By comparing the short-term power difference between the signal filtered from the microphone 4, etc. that includes the characteristics of the environment, and the digitally converted signal of the synthesized voice recorded from the output from the speaker 2 through the microphone 4, sudden noise can be detected. A means for detecting this and generating an interrupt signal (3) to the speech synthesizer 1 is a means necessary for carrying out the present invention. Note that the same reference numerals indicate the same objects throughout the figures.

The system of the embodiment (Embodiment 1) shown in FIG. 1 includes a speech synthesizer 1. Speaker 2. Microphone 4. The interrupting device 5 of the present invention includes an analog/digital converter (A/D) 51 that converts the synthesized sound (analog data) including noise extracted from the microphone 4 into digital data.
．． Analog/digital converter (A/D) that converts the synthesized sound itself (analog data) into digital data 52.
Synthetic sound signal correction unit 54. A coefficient calculation unit 539 that determines the coefficients of the signal correction unit 54; a difference signal power calculation unit 55 that compares the corrected synthesized sound with the synthesized sound from the microphone 4 that includes noise; It consists of an interruption signal generating section 56 that interrupts the synthesis when the noise is extremely large.

The signal correction unit 54 is configured to perform an FIR as shown in FIG.
The coefficients are calculated by a coefficient calculating section 53. This coefficient calculation unit 53 is the above-mentioned (0)
) solve the equation. The interruption signal generation section 56 compares the externally given threshold with the input from the difference signal power calculation section 55, and when the input is larger, sends a signal to the speech synthesizer 1.
Operates to output an interrupt signal.

The operational flow at this time is shown in FIG. First, coefficients for the signal correction section 54 are determined by the coefficient calculation section 53. This synthesizes and outputs a test message (for example, "It's a sunny day today") using the speech synthesizer W1, and uses the synthesized sound itself and the synthesized sound extracted from the microphone 4 to determine the surrounding environment at that time. (including speaker 2 and microphone 4) can be determined. After determining the coefficients, actual speech synthesis is started, and when noise is detected, the speech synthesis is interrupted. Thereafter, the speech synthesis will be resumed, for example, from the interrupted speech.

FIG. 3 shows another embodiment (Embodiment 2) of the present invention. This embodiment does not have an analog/digital converter (A/D) 52 that converts an analog signal from the speech synthesizer l into a digital signal, and detects internal data from the speech synthesizer l in the form of digital data. directly,
Signal correction section 54. The coefficient calculation unit 53 is inputted. This has the advantage that analog/digital conversion of synthesized speech requires less effort than the first embodiment shown in FIG.

FIG. 4 shows another embodiment (Embodiment 3). The coefficient calculating section 53 shown in the embodiment of FIG. Instead of the signal correction section 54, a similar coefficient calculation section 58. A signal correction section 57 is used.

This is a method of correcting the voice extracted from the microphone 4 by removing the speaker/microphone characteristics, and the synthesized voice recorded by the microphone 4 is corrected to the voice synthesizer 1 as much as possible.
This makes it easier to extract the true noise by bringing it closer to the synthesized speech itself. This is the coefficient calculating section 53 in FIG. This means that the operation is the opposite of that of the signal correction section 54.

FIG. 5 shows another embodiment (Embodiment 4). In this embodiment, like the embodiment shown in FIG. 3 (Embodiment 2), an analog/digital converter (A/
D) detects internal data from the speech synthesizer 1 in the form of digital data and inputs it to the coefficient calculation unit 58; This is a combination of Example 2 and Example 3 above.

FIG. 6 shows another embodiment (Embodiment 5). As in the first embodiment shown in FIG. 1, the synthesized signal from the speech synthesizer 1 is corrected, and as in the third embodiment shown in FIG. This is a combination of Embodiment 1 and Embodiment 3, and attempts to correct the difference, detect the difference between the two, and generate an interruption signal.

FIG. 7 shows another embodiment (Embodiment 6). In this embodiment, in the fifth embodiment, an analog/digital converter (A/D
) 52 is omitted, internal data is detected in the form of digital data from the speech synthesizer 1, and the signal correction unit 53. This is a combination of the fifth embodiment and the fourth embodiment, which is directly input to the coefficient calculating section 510.

As described above, various configurations are possible when specifically implementing the present invention.

As described above, the present invention provides a method for temporarily interrupting speech synthesis in the speech synthesizer when, while listening to synthesized speech, you suddenly cannot hear it due to ambient noise. A speaker for the output audio digital conversion signal,
Applying a filter that includes the characteristics of the surrounding environment such as the microphone,
By comparing the short-term power difference between a synthesized voice that is close to the synthesized voice recorded with a microphone and a digitally converted signal of the synthesized voice that is recorded from the output from the speaker through the microphone, we eliminate the surrounding environment conditions and suddenly detect the noise of
The feature is that the above-mentioned speech synthesis device is configured to receive an interruption signal.

〔Effect of the invention〕

As described above in detail, the speech synthesis interrupting device of the present invention suddenly interrupts voice synthesis due to ambient noise while listening to synthesized speech.
If the voice cannot be heard, - temporarily interrupting the voice synthesis, for example, in a voice synthesis device such as an accumulation synthesis (analysis synthesis) type or a rule synthesis type, the output voice from the voice synthesis device.

An acyclic type (F
IR type) filter is applied to correct the signal, the difference due to environmental conditions is minimized, and the difference is compared to detect sudden noise, and if the noise level is greater than a specific threshold, The system is configured to generate a speech synthesis interruption signal, thereby automatically interrupting speech synthesis, and then restarting speech synthesis. By automatically extracting noise different from the synthesized speech and using the extracted signal as an interruption signal, it is possible to automatically interrupt speech synthesis.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an embodiment of the present invention. FIG. 2 is an operational flow diagram of the present invention. FIGS. 3 to 7 are diagrams showing other embodiments of the present invention. FIG. 8 is a diagram showing an example of the configuration of an acyclic filter. FIG. 9 is a diagram illustrating a conventional speech synthesis interruption method. It is. In the drawing, 1 is a voice synthesizer, 2 is a speaker. 3 is the interrupt switch, 4 is the microphone. 5 is an interruption device. 51 and 52 are analog/digital converters (A/D). 53.58.510 is the coefficient calculation section. 54.57 is a signal correction section. 55 is a difference signal power calculation section, and 56 is an interruption signal generation section. ■ is an interruption signal. are shown respectively. start 1 Operation flow diagram of the present invention No. figure No. figure No. figure Diagram showing another embodiment of the present invention No. figure Diagram showing another embodiment of the present invention No. figure

Claims (5)

[Claims] (1) A microphone (4) that records the synthesized voice output from the speaker (2) of the voice synthesizer (1);
) an analog/digital converter (A/D) (51) that converts the output of the output into a digital signal, and the speech synthesizer (1)
An analog/digital converter (A/D) (
52) and the analog/digital converter (A/D) (
a signal correction section (54) that delays and amplifies (or attenuates) the output of the signal correction section (51 or 52) for a certain period of time;
), or input analog signals from the microphone (4) to an analog/digital converter (A/D) (52 or 51).
)), which calculates the short-time power of the difference between the digitally converted output and the digitally converted output; (1)) An interruption signal generation unit (56) for generating a speech synthesis interruption device. (2) The speech synthesis interrupting device according to claim 1, wherein an acyclic (FIR type) filter is used as the signal correction section (54). (3) In the speech synthesis interrupting device, the analog/digital converter (A/D) (
52), and the digital input to the digital-to-analog converter (D/A) in the speech synthesizer (1) is directly passed to the signal correction unit (54). The speech synthesis interrupting device according to item 1. (4) In the above speech synthesis interrupting device, each analog/
Claims 1 and 2 further comprising a coefficient calculation unit (53) that calculates optimal coefficients of the acyclic (FIR type) filter from the output of the digital conversion unit (A/D) (51, 52). The device for interrupting the speech synthesis described above. (5) In the above-mentioned speech synthesis interrupting device, instead of the signal correction section that corrects the output from the speech synthesis device (1), the signal correction section delays and amplifies (or attenuates) the output of the microphone (4) for a certain period of time. The speech synthesis interrupting device according to claim 1, characterized in that the device uses a section (57).