JPH05307395A - Voice synthesizer - Google Patents

Abstract

PURPOSE:To obtain synthesized voices which are easily recognized in a varying environmental noise condition. CONSTITUTION:The synthesizer consists of a voice synthesizer signal generating circuit 20, a speaker 6 which transforms synthesized voice signals into sound waves and outputs, a microphone 16 which takes in the synthesized voices as well as surrounding noises and transforms them and a surrounding sound extraction and analyzing circuit 21 extracts and analyzes the surrounding noise components of the voice signals taken in by the microphone 16. The parameters used to generate the synthesized voice signals are adjusted based on the result of the analysis.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech synthesizer that artificially produces speech.

[0002]

2. Description of the Related Art As a recent voice synthesizing method for artificially producing a voice, for example, a recording / editing method in which a voice waveform uttered by a person is stored as it is or after waveform encoding is stored and connected as needed to be output. , A parameter editing method that analyzes the voice waveform uttered by a person, accumulates it in the form converted into parameters, and connects it to drive a voice synthesizer to produce a voice, and a character string or phoneme symbol string There is a rule synthesizing method that produces speech based on phonetic and linguistic rules.

That is, in the voice synthesis by the above-mentioned recording / editing method, the voice uttered by a person is accumulated (recorded) in units of words, phrases, etc., and read out and connected (edited) as needed. , A method of synthesizing voice.
Generally, in the recording / editing method, the quality of the text voice depends on the continuity of the acoustic characteristics (spectrum envelope, amplitude, fundamental frequency, utterance speed, etc.) of the voice in the connection of different storage unit times. The larger the unit to be accumulated, such as bunsetsu and sentence, the better the quality of the produced speech (clarity, naturalness, etc.), but the vocabulary and types of sentences that can be synthesized are limited. On the other hand, if the storage unit is reduced to a syllable or a single sound, the degree of freedom of vocabulary and sentences that can be synthesized is increased, but the quality of speech is significantly reduced.

Further, the speech synthesis by the parameter editing method uses words, phrases, etc. as a unit as in the case of the recording and editing method, but the speech uttered by a person is analyzed in advance based on the speech generation model, Store in the form of parameter time series,
In this method, a voice synthesizer is driven by using a parameter time series connected as needed to synthesize voice. Since the sound source and the spectrum envelope parameter are stored, the naturalness of the synthesized speech is slightly lowered depending on the method as compared with the case where the waveform is stored, but a large amount of information compression can be achieved. Further, by manipulating the parameters, it is possible to expand / contract the time length, smooth the pitch of the connecting portion, or change the spectrum.

Further, in the speech synthesis by the rule synthesis method, characteristic parameters are used in basic small units such as syllables, phonemes, and waveforms in one pitch interval as a unit to be stored, and instead, rules for connecting them are used. In addition, it is a method that allows any word to be synthesized from a sequence of phonemes, syllable symbols, or characters by precisely defining rules for controlling prosodic information such as pitch and amplitude. At this time, in order for the voice to be natural and easy to hear, changes in pitch and stress, temporal changes in the spectrum, and poses must be natural. Therefore,
In the case of this method, it is based on the phonetic or linguistic characteristics of natural speech together with the quality of the basic unit used for synthesis.
The control rules (control information and control mechanism) of acoustic parameters play an important role.

FIG. 8 shows an example of a conventional speech synthesizer that artificially synthesizes speech as described above. In FIG. 8, the sound source generation unit 2 generates sound source data based on the sound source parameters supplied from the sound source parameter storage unit 7. This sound source data is used as the voice synthesis filter 3
Sent to. The speech synthesis filter 3 is also supplied with the spectrum parameter from the spectrum parameter storage unit 8. Therefore, the speech synthesis filter 3 synthesizes the sound source data by filtering the sound source data based on the spectrum parameter. The waveform data of the audio signal can be obtained. The waveform data of the synthesized voice signal is sent to the level adjuster 4. The level adjuster 4 is also supplied with level adjustment data from the volume control unit 1 corresponding to the volume set by the user. Therefore, based on the level adjustment data set by the user with the level adjuster 4, The level of the synthesized voice signal is adjusted. The synthesized voice signal whose level has been adjusted by the level adjuster 4 is converted into an analog synthesized voice signal by the digital / analog (D / A) converter 5, and further output as synthesized voice by the sound emitting means such as the speaker 6. To be done.

[0007]

By the way, since the above-mentioned conventional speech synthesizer is premised on the case where it is placed in a constant acoustic environment as a use environment, the speech synthesis output level is usually set to a constant value. Is set to.

However, in the conventional speech synthesizer as described above, the following problems occur when the surrounding acoustic environment changes.

That is, as a case where the surrounding acoustic environment changes, for example, a case where a voice synthesizer is used in a mobile robot or in a car is considered. When the surrounding acoustic environment changes in this way, it becomes difficult to hear the synthesized voice. Therefore, in order to make it easier to hear the synthesized voice, the user who uses the conventional speech synthesizer adjusts the volume according to the surrounding acoustic environment, uses headphones, or once again makes a speech. You have to do things such as redoing (rewording) the composition. This is not an easy-to-use device. As described above, the conventional speech synthesizer has a problem in that sufficient consideration is not taken into consideration when the noise environment in which the speech synthesizer is used changes.

Therefore, the present invention has been proposed in view of the above-mentioned circumstances, and it is possible to obtain a synthesized speech that is easy to hear even in a situation where the ambient noise environment changes. The purpose is to provide a device.

[0011]

The speech synthesizer of the present invention is proposed to achieve the above-mentioned object, and includes a synthesized speech signal generation means for generating a synthesized speech signal using a predetermined parameter. A sound emitting means for converting the synthesized speech signal from the synthesized speech signal generating means into a sound wave and outputting the synthesized speech; and a collection means for capturing ambient noise together with the synthesized speech output from the sound emitting means and converting it into a speech signal. It has a sounding means and an analyzing means for extracting the ambient noise component of the voice signal taken in by the sound collecting means and analyzing the ambient noise component, and the analysis result by the analyzing means (the analysis result of the ambient noise component). On the basis of the above, the above-mentioned parameter is adjusted when the synthesized voice signal is generated by the synthesized voice signal generating means.

That is, the speech synthesizer of the present invention is a speech synthesizer used in an environment in which the ambient noise environment changes, for example. It is designed to obtain a voice synthesis output that is easy to hear by taking in the device by using sound collecting means such as etc., analyzing the ambient noise, and adjusting the parameters when synthesizing the voice based on the analysis result. ..

Further, the speech synthesizing apparatus of the present invention generates a synthesized speech signal using the sound source parameter and the spectrum parameter, adjusts the level and outputs the synthesized speech signal, and the synthesized speech signal generating means. Sound emitting means for converting a synthetic voice signal into a sound wave and outputting a synthetic voice,
A sound collecting unit that takes in ambient noise together with the synthesized voice output from the sound emitting unit and converts the ambient noise into a voice signal, and extracts a ambient noise component of the voice signal captured by the sound collecting unit to extract the ambient noise component by a fast Fourier transform. And an analysis unit for performing frequency analysis by transform (FFT) processing and calculating a so-called masking curve when another sound is masked by another sound based on the frequency analysis result (frequency analysis result of ambient noise component). Using the masking curve calculated by the analyzing means, the sound source parameter, the spectrum parameter, and the level adjustment value in the synthetic voice signal generating means are controlled so that the synthetic voice output from the sound emitting means is not masked by the ambient noise. It is designed to control.

That is, this speech synthesizer performs a frequency analysis by applying a fast Fourier transform (FFT) to the ambient noise signal, obtains a masking curve from the spectrum of the result of this frequency analysis, and outputs using this masking curve. The sound source parameter, the spectrum parameter, and the level adjustment value are controlled so that the synthesized sound is not masked by the ambient sound.

[0015]

According to the voice synthesizing apparatus of the present invention, the synthesized voice once synthesized and output from the sound emitting means is taken in together with the ambient noise by the sound collecting means, and the ambient noise signal is extracted from the signal taken in by the sound collecting means. Then, the parameters are adjusted based on the extraction and analysis result of the ambient noise signal, and the voice is synthesized again to obtain a synthesized voice that is not easily influenced by the ambient noise.

[0016]

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

As shown in FIG. 1, the basic configuration of the speech synthesizing apparatus according to the present invention comprises a synthetic speech signal generating circuit 20 for generating a synthetic speech signal with a predetermined parameter, and the synthetic speech signal generating circuit 20. A sound emitting means (speaker 6) for converting a synthetic voice signal into a sound wave and outputting a synthetic voice, and a sound collecting means for capturing ambient noise together with the synthetic voice output from the speaker 6 and converting it into a voice signal (microphone 1
6) and an analyzing means (ambient sound extraction / analysis circuit 21) for extracting an ambient noise component of the voice signal captured by the microphone 16 and analyzing the extracted ambient noise component. The above parameters are adjusted when the synthesized voice signal is generated by the synthesized voice signal generation circuit 20 based on the analysis result in.

That is, the voice synthesizing apparatus of this embodiment is constructed so as to obtain a synthesized voice output that is easy to hear even in an acoustic environment in which the surrounding acoustic environment changes as described above. The sound produced by the synthesized voice signal generation circuit 20 and output from the speaker 6 together with the current surrounding acoustic environment is taken into the device by the microphone 16, and the sound taken by the microphone 16 by the ambient sound extraction / analysis circuit 21. A noise (ambient noise) component is extracted from the noise component, the noise component is analyzed, and the synthesized speech signal generation circuit 20 adjusts the speech synthesis parameter based on the analysis result so that the synthesized speech can be easily heard. Have control. In addition,
In the present embodiment, since the signals are handled digitally, the above-mentioned synthesized voice signal which is digital data is converted into an analog signal through the digital / analog (D / A) converter 5 and then the above-mentioned speaker. 6, the analog voice signal from the microphone 16 is converted into a digital signal via the analog / digital (A / D) converter 15 and is sent to the ambient sound extraction / analysis circuit 21. There is.

FIG. 2 shows the structure of a more detailed specific example (first specific example) of the above basic structure. In FIG. 2, the same components as those in FIG. 1 and FIG. 8 described above are designated by the same reference numerals.

Here, the synthesized voice signal generation circuit 20 of the first concrete example includes a volume control unit 1, a sound source generation unit 2, a voice synthesis filter 3, a sound source parameter storage unit 7, a spectrum parameter storage unit 8 and a level. The ambient sound extraction / analysis circuit (analyzing means) 21 includes the adjuster 10.
Is composed of a buffer 11, an RMS (root mean square) value calculation unit 12, an adder 13, and a buffer 14.

That is, in FIG. 2, the sound source generation section 2 generates sound source data based on the sound source parameters supplied from the sound source parameter storage section 7. This sound source data is sent to the voice synthesis filter 3. The speech synthesis filter 3 is also supplied with the spectrum parameter from the spectrum parameter storage unit 8. Therefore, the speech synthesis filter 3 synthesizes the sound source data by filtering the sound source data based on the spectrum parameter. The waveform data of the audio signal can be obtained. The waveform data of the synthesized voice signal is sent to the level adjuster 10. This level adjuster 10
Is also supplied with level adjustment data from the volume control unit 1 corresponding to the volume set by the user, and therefore,
The level adjuster 10 adjusts the level of the synthesized voice signal based on the level adjustment data set by the user. The synthesized voice signal whose level is adjusted by the level adjuster 10 is converted into an analog synthesized voice signal by the D / A converter 5 and is further output as a synthesized voice by the sound emitting means such as the speaker 6.

On the other hand, in the microphone 16, ambient noise is also captured together with the synthesized voice output from the speaker 6. The audio signal taken in by the microphone 16 is converted into a digital audio signal via the A / D converter 15, and then temporarily stored in the buffer 14 and then sent to the adder 13 as an addition signal. The adder 13 is supplied with the synthesized voice signal output from the level adjuster 10 via the buffer 11 as a subtraction signal. The buffers 11 and 14 are
Is the level adjuster 10 supplied to the adder 13.
The synthesized voice signal output from the microphone 16
It is provided in order to align the timing so that the synthesized voice signal included in the voice signal captured by the input A / D converter 15 is the same. As a result, the output of the adder 15 becomes a signal obtained by removing the synthesized voice signal component taken in together with the ambient noise by the microphone 16, that is, a signal of only the ambient noise component. The signal of the ambient noise component from the adder 15 is the RMS
The value is sent to the value calculator 12, and the calculator 12 calculates the RMS value. The RMS value is sent to the level adjuster 10, and the level adjuster 10 adjusts the level based on the RMS value.

In other words, in the configuration of the first specific example, the sound (sound including synthetic speech and ambient noise) under the speech synthesis output environment is captured by the microphone 16 and A / D converted to the buffer 14. Stored, the waveform data of the synthesized voice signal output at that time is also stored in the buffer 11, and the noise (ambient noise) component is extracted by the adder 13 from the waveform data stored in both the buffers 11 and 14. There is. Using this noise component, the RMS value calculation unit 12 obtains the RMS value of the ambient noise, inputs this RMS value to the level adjuster 10 that performs level adjustment on the speech synthesis output, and sets the speech synthesis output level to the ambient sound. Control and output so that the sound becomes louder.

Thus, according to the configuration of the first specific example, as shown in FIG. 3, the preset volume (set volume) of the output volume (level) A _OUT of the synthesized voice from the speaker 6 is set. Ambient noise level (ambient sound RMS
When the value (level corresponding to the value) a _rms is high, the output volume A _OUT can be controlled to be louder than the ambient sound and output. The output volume A
_{The level at which OUT} is controlled to be louder than ambient sound can be preset by the user.

Next, the speech synthesizer of the present invention shown in FIG. 1 can be configured as in the second concrete example shown in FIG. Note that, in FIG. 4, the same reference numerals are given to constituent elements corresponding to those in FIGS. 1 and 2 described above.

The synthesized voice signal generation circuit 20 of the second specific example device shown in FIG. 4 includes the volume control unit 1, the sound source generation unit 2, the voice synthesis filter 3, the sound source parameter storage unit 30, and the spectrum parameter storage unit 31. And a level adjuster 32 to generate a synthesized voice signal using the sound source parameter and the spectrum parameter and adjust and output the level. The ambient sound extraction / analysis circuit (analyzing means) 21 includes the buffers 11 and 14
And an adder 13, a fast Fourier transform (FFT) circuit 34, and a masking curve calculator 33. The ambient noise component of the voice signal captured by the microphone 16 is extracted, and the extracted ambient noise component is subjected to the fast Fourier transform. Frequency analysis is performed by the fast Fourier transform processing in the unit 34, and the masking curve calculation unit 33 calculates the masking curve based on the frequency analysis result (frequency analysis result of the ambient noise component).

In the second specific configuration, the synthetic voice output from the speaker 6 using the masking curve data calculated by the masking curve calculating unit 33 is prevented from being masked by the ambient noise. The sound source parameter, the spectrum parameter, and the level adjustment value in the signal generation circuit 20 are controlled.

That is, in FIG. 4, the sound source generation section 2 generates sound source data based on the sound source parameters supplied from the sound source parameter storage section 30, and the sound source data is sent to the voice synthesis filter 3. A spectrum parameter from the spectrum parameter accumulating unit 31 is also supplied to the speech synthesis filter 3, and the speech synthesis filter 3 filters the sound source data based on the spectrum parameter to synthesize a speech signal. Waveform data is obtained. The waveform data of the synthesized voice signal is sent to the level adjuster 32, and the level adjuster 10 causes the volume controller 1 to output the volume data.
The level of the synthesized voice signal is adjusted based on the level adjustment data from. The synthesized speech signal whose level is adjusted by the level adjuster 32 is output as synthesized speech by the speaker 6 via the D / A converter 5.

On the other hand, the microphone 16 also takes in ambient noise together with the synthesized voice output from the speaker 6,
The audio signal captured by this microphone 16 is A / D
After being temporarily stored in the buffer 14 via the converter 15, it is sent to the adder 13 as an addition signal. Also,
The adder 13 is supplied with the synthetic speech signal output from the level adjuster 32 and passed through the buffer 11 as a subtraction signal. The output of the adder 15 is a signal obtained by removing the synthesized voice signal component taken in together with the ambient noise by the microphone 16 as described above, that is, a signal of only the ambient noise component. The signal of the ambient noise component from the adder 15 is sent to the fast Fourier transform unit 34, and the fast Fourier transform unit 34 performs fast Fourier transform processing on the signal. This fast Fourier transform processing result (spectral data) is sent to the masking curve calculation unit 33, and in the masking curve calculation unit 33, a certain sound (ambient noise) masks another sound (synthesized voice), so-called masking. A curve is required. The masking curve data is sent to the sound source parameter accumulating unit 30, the spectrum parameter accumulating unit 31, and the level adjuster 32, and these accumulating units 30, 31 and the level adjuster 32.
Then, the masking curve data is used to control the sound source parameter, the spectrum parameter, and the level adjustment value so that the output synthesized sound is not masked by ambient noise.

Here, an example of the parameter adjustment in the second specific example will be shown below. For example, when the ambient noise has a waveform like a sine wave, as shown in FIG.
The spectral parameters are changed so as to avoid the frequency of the sine wave (the frequency of ambient noise) so that the so-called formant frequency corresponding to the resonance frequency of the vocal tract is changed. In FIG. 5, the curve EN _A in the figure shows the envelope of the synthetic speech, the curve MC _{a in} the figure is the masking curve of the ambient noise, and the curve en _{A in} the figure is the envelope of the synthetic speech corrected by this example. Indicates. Similarly, when the ambient noise has a waveform like a sine wave, for example, as shown in FIG.
As shown in, the sound source parameter is changed to change the pitch frequency. In addition, in FIG.
In the figure, the curve PT _A indicates the pitch of the synthesized voice, and the curve MC _a
Indicates the masking curve of ambient noise, and the curve pt _{A in} the figure indicates the pitch of the synthesized speech corrected by this example.

Further, when the ambient noise is band noise, as shown in FIG. 7, the spectrum parameter is changed to adjust the frequency envelope level so as not to fall below the masking curve. Do things. In FIG. 7, the curve EN _A in the figure shows the envelope of the synthetic speech, the curve MC _{a in} the figure is the masking curve of the ambient noise, and the curve en _{A in} the figure is the envelope of the synthetic speech corrected by this specific example. Indicates.

From the above, according to the speech synthesizer of the present embodiment, it is difficult for the conventional speech synthesizer to comprehend in a situation where the ambient noise environment changes (in a mobile robot, an automobile, etc.). It is possible to obtain a synthetic voice output that is easier to hear than voice output.

[0033]

As described above, in the voice synthesizing apparatus of the present invention, a synthesized voice signal is generated using a predetermined parameter, and the synthesized voice signal is converted into a sound wave and output, and ambient noise is generated. A situation in which the ambient noise environment changes by adjusting the parameters when capturing and converting to a voice signal and analyzing the ambient noise component of this voice signal based on the analysis results. Even below, it is possible to obtain a synthesized voice that is easy to hear.

[Brief description of drawings]

FIG. 1 is a block circuit diagram showing a basic configuration of a speech synthesizer according to an embodiment of the present invention.

FIG. 2 is a block circuit diagram showing a first specific configuration of the present embodiment.

FIG. 3 is a diagram for explaining a state of level adjustment in the first specific example.

FIG. 4 is a block circuit diagram showing a second specific configuration of the present embodiment.

FIG. 5 is a diagram for explaining how the spectrum parameters are adjusted when the ambient noise is a sine wave in the second specific example.

FIG. 6 is a diagram for explaining how the sound source parameters are adjusted when the ambient noise is a sine wave in the second specific example.

FIG. 7 is a diagram for explaining how the spectrum parameter is adjusted when the ambient noise is band noise in the second specific example.

FIG. 8 is a block circuit diagram showing a schematic configuration of a conventional speech synthesizer.

[Explanation of symbols]

 1 ... Volume control unit 2 ... Sound source generation unit 3 ... Voice synthesis filter 5 ... D / A converter 6 ... Speaker 7, 30 ... ..Sound source parameter storage unit 8,31 ... Spectral parameter storage unit 10,32 ... Level adjuster 11,14 ... Buffer 12 ... RMS value calculation unit 13 ... Adder 15・ ・ ・ A / D converter 20 ・ ・ ・ Synthetic voice signal generation circuit 21 ・ ・ ・ ・ Ambient sound extraction analysis circuit 33 ・ ・ ・ ・ Masking curve calculation unit 34 ・ ・ ・ ・ ・ ・ Fast Fourier transform unit

Claims (2)

[Claims] 1. Synthetic voice signal generating means for generating a synthetic voice signal using a predetermined parameter, and sound emitting means for converting the synthetic voice signal from the synthetic voice signal generating means into sound waves to output a synthetic voice. A sound collecting unit that captures an ambient sound together with the synthesized voice output from the sound emitting unit and converts the ambient sound into an audio signal; and an analyzing unit that analyzes an ambient sound component of the audio signal captured by the sound collecting unit, A voice synthesizing apparatus, characterized in that the above-mentioned parameter is adjusted when the synthetic voice signal is generated by the synthetic voice signal generating means based on the analysis result of the analyzing means. 2. A synthetic speech signal generating means for generating a synthetic speech signal using a sound source parameter and a spectrum parameter and adjusting and outputting a level, and converting the synthetic speech signal from the synthetic speech signal generating means into a sound wave. And a sound collecting means for outputting a synthesized voice, a sound collecting means for collecting a surrounding sound together with the synthesized sound outputted from the sound emitting means and converting the sound into a sound signal, and an ambient sound component of the sound signal fetched by the sound collecting means. And a analyzing means for calculating a masking curve on the basis of the frequency analysis result by the fast Fourier transform processing, and using the masking curve calculated by the analyzing means, the synthesis output from the sound emitting means. In order to prevent the voice from being masked by the ambient sound, the sound source parameter, the spectrum parameter and the level in the synthesized voice signal generating means are set. Speech synthesis apparatus characterized by controlling the adjustment value.