JPH07191686A - Device for reading document aloud - Google Patents

Abstract

PURPOSE:To provide the document reading device which is improved in operation convenience when a document reading board is mounted on a computer in which a voice reproducing function is incorporated. CONSTITUTION:A CPU 1, a voice synthesizer 5, an oversampling digital filter 5 for return noise removal of sampling, and a 1-bit D/A converter 7 which converts digital voice data into a binary pulse train by SIGMA-DELTA conversion are integrated on one chip; and the binary pulse train which is generated here is transmitted through an IC card slot as the interface of the computer to provide a function which reads text data aloud through an LPF, an amplifier, and a loudspeaker incorporated in the computer main body.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an improvement of a text-to-speech device for directly converting a mixed kanji / kana sentence into a voice.

[0002]

2. Description of the Related Art A text-to-speech device has been developed in which a voice signal is synthesized and output based on a text containing kanji and kana created by text-writing software or text-editing software (text editor) and the voice is read aloud.

This text-to-speech device directly produces natural voice signals from ordinary text data.
It can be used for various purposes because you can directly listen to mixed kanji and kana sentences.

The text-to-speech device is realized as a text-to-speech synthesis board (option board) in order to downsize the system and reduce the load on the host computer. That is, a personal computer (PC) or workstation (W
S), a board mounted in an expansion slot of a host computer such as a word processor (WP), or an IC
It is provided in the form of an IC card mounted in a card slot. FIG. 5 shows its internal configuration.

The configuration operation will be briefly described below with reference to FIG. The text data is generated by the host computer and sent to the text reading device via the connector. The text-to-speech device performs, on the text data generated by the host computer, linguistic processing such as word division and accent information addition by the linguistic processing device 51 to read out a speech symbol string (prosodic information and phonological sequence). obtain. The voice symbol string is sent to the voice synthesizer 52, where the voice unit is extracted and the voice signal is regularly synthesized. The voice signal is amplified by the analog amplifier 53, and the voice can be heard by an external earphone or speaker 54.

[0006]

By the way, a personal computer (PC), a workstation (WS),
Some host computers, such as word processors (WP), have a built-in circuit for realizing voice recording and reproduction, including an amplifier and a speaker, in the main body. FIG. 6 shows an example of the configuration.

In FIG. 6, reference numeral 60 is a main body (referred to as a PC / WS / WP main body) of a personal computer (PC), workstation (WS), word processor (WP), etc., and reference numeral 61 is the configuration shown in FIG. The text reading device, 62, 63, 64, 65 are PCs, respectively
/ WS / WP body 60 is built in, 62 is a body signal source, 63 is an amplifier, 64 is a low-pass filter (LP
F) and 65 are speakers.

On the other hand, since a general-purpose expansion slot or IC card slot equipped in a host computer such as a PC / WS / WP has an interface for digital signal processing, an analog voice signal generated by a text-to-speech device is required. Cannot be transferred to the host computer main body via the expansion slot or IC card slot.

Therefore, the read-out voice synthesized by the text-to-speech device is used by the speaker 6 of the host computer.
When you want to play and listen with 5, you can use a dedicated analog signal line (connection cable 68, analog signal input connector 69) that is different from the general-purpose expansion slot or IC card slot.
It was necessary to provide an analog signal from the text reading device to the host computer body.

In this case, since a connection cable is required externally, the connection is troublesome, troublesome and difficult to use. There is also a problem that a dedicated analog voice input interface is required on the main body side.

The present invention has been made in view of the above circumstances, and it is an object of the present invention to provide a text-to-speech device capable of improving usability when a text-to-speech board is mounted on a host computer having a voice reproduction function. To aim.

[0012]

A text-to-speech device of the present invention comprises a computer main body for generating text data,
A sentence-speech synthesizer for linguistically analyzing text data generated by a computer body and converting it into digital voice data, and a 1-bit D / A for converting the digital voice data into energy density wave voice data by a binary pulse train.
2 converted by a converter and a 1-bit D / A converter
And a data transfer device for outputting a pulse train of values via a digital interface. The computer has means for audio reproduction including an amplifier and a speaker. Further, the sentence-speech synthesizer performs a language process for reading a text data sent from a computer according to a program stored in a built-in memory, while referring to a built-in language processing dictionary to generate a voice symbol string. Then, it is generated by the speech processing circuit that extracts the speech unit from the speech symbol string by referring to the built-in speech unit data, extracts the necessary parameters, and outputs the parameters to the speech synthesizer, and the speech processing circuit. A voice synthesizer that processes digital parameters to convert the digital voice data to digital voice data, and the digital voice data generated by the voice generator is oversampled to interpolate between sampling points and remove aliasing noise included in the voice data. And a digital filter that performs Σ-Δ conversion to convert the digital audio data into a binary pulse train. 1-bit D / A to
A converter, a data transfer device for outputting a binary pulse train converted by the 1-bit D / A converter through a digital interface, and a binary pulse train converted by the 1-bit D / A converter for analog voice. It is characterized by including a low-pass filter for converting into a signal, an amplifier, and an analog signal output connector. Further, the voice processing circuit, the voice synthesizer, the digital filter and the 1-bit D / A converter are integrated in a 1-chip LSI, and the digital voice data generated by the 1-bit D / A converter is interfaced according to a certain industrial standard. It is also characterized in that it is supplied to a computer having a retransmission function via.

[0013]

The text-to-speech device of the present invention comprises a CPU, a voice synthesizer, an oversampling digital filter for removing sampling aliasing noise, and a 1-bit converting digital voice data into a binary pulse train by Σ-Δ conversion. D / A converter and 1 chip integrated, and CPU
RAM for storing programs for operating the computer, language processing dictionaries, and voice unit data for voice synthesis
A RAM for storing text data; an IC card slot for interfacing with other information processing devices such as PC / WS; an LPF (low-pass filter) for converting a binary pulse train into an analog audio signal; and an amplifier. The earphone jack is stored in a credit card-sized IC card, and the IC card realizes a function of reading aloud text data sent from a computer. Specifically, C
The PU performs a language processing for reading the text data sent from the computer according to the program stored in the built-in memory while referring to the built-in language processing dictionary to obtain a phonetic symbol string, and further The speech unit is extracted from the speech symbol string by referring to the speech unit data, necessary parameters are extracted, and the parameters are output to the speech synthesizer. The voice synthesizer digitally processes the parameters generated by the CPU to convert them into digital voicing data, oversamples the digital voice data by a digital filter, interpolates between sampling points, and returns the data included in the voice data. Remove noise. Then, the digital audio data is converted into a binary pulse train by Σ-Δ conversion.

Further, the binary audio signal converted by the D / A converter is transferred to a PC / WP or the like via a digital signal interface of an IC card slot interface, and the binary audio signal is represented by PC / WS. Low pass filter (LP
F), which provides a function of reproducing with a speaker through an amplifier.

According to the present invention, when the voice reproduced by the text-to-speech device is reproduced by the speaker built in the computer body, it is transferred through the digital interface of the general-purpose IC card interface, and the computer body is a simple LPF. Since analog audio can be played back only by passing it through, only a simple LPF is required on the computer side to obtain an analog audio signal. Therefore, no external connection cable is required and no dedicated analog audio input interface is required. It has the advantage of becoming. Also, I
Since the audio signal transferred from the C card interface is binary digital data, it is possible to obtain an effect that control on the main body side becomes easy.

[0016]

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

FIG. 1 is a block diagram showing an embodiment of the present invention.

In the figure, reference numeral 1 is a CPU for controlling the reading of a document according to the present invention. Reference numeral 2 is a ROM, C
A program and data a for operating the PU1, a dictionary b for language processing, and speech unit data c for speech synthesis.
Is stored.

Reference numeral 3 is a work RAM, which is a dynamic R in which the CPU 1 temporarily stores various data being processed.
AM. Reference numeral 4 is a RAM for storing document data,
This is a static RAM that stores text data (text data) sent through an external interface and is backed up by a battery.

Reference numeral 5 is a voice synthesizer, which digitally processes the data from the CPU 1 to convert it into digital voicing data (PCM voice data). Reference numeral 6 is an oversampling digital filter that removes aliasing noise due to sampling by oversampling the digital voice data synthesized by the voice synthesizer 5.

Reference numeral 7 is a 1-bit D / A converter, which further oversamples the digital audio data from the digital filter and modulates it into a binary PWM (Pulse Width Modulation) signal by the Σ-Δ conversion method.

Reference numeral 8 is an LPF (low pass filter), which is a PW generated by the 1-bit D / A converter 7.
An analog signal is generated from the M signal. Reference numeral 9 is an amplifier, which amplifies the analog audio signal by voltage / power.
Reference numeral 10 is an earphone jack, which outputs an analog audio signal to the outside. You can listen to the read voice by connecting an earphone directly to this earphone jack or connecting a speaker with a built-in power amplifier.

Reference numeral 11 is an internal data bus. Reference numeral 12 is an external interface, for example, JEIDA
It is a standard interface for connecting an IC card conforming to the version 4 standard and an information processing device such as a PC or WS. In the figure, CPU 1, voice synthesizer 5, oversampling digital filter 6, and 1-bit D / A
The converter 7 is mounted on a 1-chip LSI.

FIGS. 2 and 3 are respectively cited for explaining the operation of the embodiment of the present invention, and FIG. 2 is a diagram for explaining oversampling and interpolation by an oversampling digital filter, and FIG. [Fig. 6] is a diagram showing an example of data representation by a PWM wave.

FIG. 4 is a PC / WS / W to which the present invention is applied.
It is a figure which shows the internal structure of a host computer, such as P.
It is shown in the sense of comparison and control with the conventional example shown in FIG. According to the present invention, a dedicated analog signal line (connection cable 68, analog signal input connector 6) shown as a conventional example in FIG.
It shows that 9) becomes unnecessary. In addition, in FIG.
The blocks with the same numbers as in FIG. 6 are the same as those shown in FIG.

The operation of the embodiment of the present invention shown in FIG. 1 will be described below with reference to FIGS. 2 and 3.

First, text data (text data) is generated by a host computer such as a PC / WS / WP and sent to the text reading device of the present invention through the external interface 12. In the text reading device, CPU1
Operates according to the program 2a stored in the ROM 2 and performs language processing for reading the sent text data such as word division and accent information addition while referring to the built-in language processing dictionary 2b, and outputs the voice. Obtain a symbol string (prosodic information and phoneme sequence).

The CPU further refers to the speech unit data 2c, extracts a speech unit from the speech symbol string, extracts a necessary parameter, and sends the parameter to the speech synthesizer 5.

The voice synthesizer 5 processes the parameters sent from the CPU 1 into digital signals and converts them into digital voicing data (PCM voice data) having a sampling frequency (fs) = 12 KHz and a quantization precision = 12 bits.

The PCM voice data generated by the voice synthesizer 5 is the oversampling digital filter 6
Sampling frequency fs by 8 times (8fs = 96K
Hz) to interpolate between sampling points to remove aliasing noise included in the PCM audio data. FIG. 2 shows an example of the oversampling and interpolation. In the figure, the circles indicate the original sampling points,
A cross mark indicates an interpolated sampling point.

Oversampling Digital Filter 6
The 1-bit D / A converter 7 performs Σ-Δ conversion on the PCM audio data oversampled by the oversampling, the quantization number reduction, and the noise shaping, and the sampling frequency 192fs
Convert to PWM waveform of (= 2.304 MHz).

The PWM waveform is either High or Low 2
A value pulse train, in which the level is expressed by the pulse width. That is, the wider the pulse width is, the higher the level is, and the narrower the pulse width is, the smaller the level is. FIG. 7 shows an example of data representation by a PWM wave. The PWM audio data generated by the 1-bit D / A converter 7 is converted into an analog audio waveform through the analog LPF 8. The analog voice waveform generated here is amplified by the amplifier 9 and output to the outside world via the earphone jack 10.

PWM generated by the digital filter 6
The waveform is sent to a host computer such as a PC / WS / WP via a digital signal interface 12 serving as an IC card interface, where the low pass filter 8
Is converted into an analog waveform through the same built-in component as above, amplified by the built-in amplifier 9 and reproduced through the speaker.

The oversampling digital filter 6 and the 1-bit D / A converter 7 used in the embodiment shown in FIG. 1 are components frequently used in recent years for reproducing digital audio of digital audio equipment such as CD players. In the embodiment of the present invention, the 1-bit D / A converter is used to output the PWM waveform. However, another method of converting digital voice data into voice data of energy density wave by a binary pulse train For example, a 1-bit D / A converter that outputs a PDM (Pulse Density Modulation) waveform may be used.

[0035]

As described above, according to the present invention, a general-purpose IC card is used when the voice reproduced by the text-to-speech device is reproduced by the speaker built in the host computer such as PC / WS / WP. Since it can be transferred to the host computer via the digital interface that serves as the interface and the main body side can reproduce the analog voice only by passing it through a simple low-pass filter, the effects listed below can be obtained. (1) An external connection cable is not required when a reading voice is reproduced by a speaker built in a host computer such as a PC / WS / WP, and usability is improved. (2) A dedicated interface for analog voice input on the host computer body side such as PC / WS / WP is not required. (3) A simple low-pass filter is required on the host computer main body side to obtain an analog audio signal. (4) Since the audio signal transferred from the IC card interface is binary digital data, it is easy to control on the main body side.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a diagram for explaining oversampling and interpolation by an oversampling digital filter.

FIG. 3 is a diagram showing an example of data representation by a PWM wave.

FIG. 4 is a diagram showing a configuration of a text-to-speech device of the present invention.

FIG. 5 is a diagram showing an internal configuration of a text-to-speech device.

FIG. 6 is a diagram showing a configuration of a conventional text-to-speech device.

[Explanation of symbols]

1 ... CPU (voice processing circuit), 2 ... ROM, 3 ... Work RAM, 4 ... SRAM, 5 ... Voice synthesizer, 6 ... Digital filter, 7 ... 1-bit D / A converter, 8 ... Low-pass filter, 9 ... Amplifier 10 ... Earphone, 11 ... Internal bus, 12 ... IC card interface.

Claims (4)

[Claims] 1. A computer main body for generating text data, a sentence / speech synthesizer for linguistically analyzing the text data generated by the computer main body for conversion into digital voice data, and an energy density of the digital voice data by a binary pulse train. A 1-bit D / A converter for converting into wave sound data, and a data transfer device for outputting a binary pulse train converted by the 1-bit D / A converter through a digital interface. Sentence reading device. 2. The text-to-speech device according to claim 1, wherein the computer has means for reproducing sound including an amplifier and a speaker. 3. The sentence voice synthesizing device performs language processing for reading aloud text data sent from a computer according to a program stored in a built-in memory, with reference to a built-in language processing dictionary. A speech processing circuit that obtains a speech symbol string, extracts the speech segment from the speech symbol string by referring to the built-in speech segment data, extracts the necessary parameters, and outputs the parameters to the speech synthesizer. A voice synthesizer that performs digital signal processing on the parameters generated by the processing circuit to convert the digital voice data, and digital voice data generated by the voice generator is oversampled to interpolate between sampling points and included in the voice data. Digital audio data is removed by Σ-Δ conversion and a digital filter that removes aliasing noise. 1 bit D converting the pulse train
/ A converter, a data transfer device for outputting the binary pulse train converted by the 1-bit D / A converter through a digital interface, and the binary pulse train converted by the 1-bit D / A converter. The text-to-speech device according to claim 1, further comprising a low-pass filter for converting into an analog voice signal, an amplifier, and an analog signal output connector. 4. The voice processing circuit, voice synthesizer, digital filter and 1-bit D / A converter are
4. The text reading system according to claim 3, wherein digital voice data integrated in a chip LSI and generated by a 1-bit D / A converter is supplied to a computer having a resending function through an interface conforming to a certain industry standard. apparatus.