JPS5853360B2 - voice response device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a voice response device that is installed as part of a telephone service system and is capable of transmitting arbitrary words.

A word-editing voice response device is a device that automatically responds by voice to a nearly limitless number of names of people, places, etc. at the request of a telephone subscriber, and it generates the requested voice by combining phonemes or phoneme chains with it. There are so-called composite voice response devices that have connected devices.

The former word editing type voice response device divides speech into words and stores them in a memory, and sends the speech by editing the words.

A device that generates speech by combining phonemes or phoneme chains performs internal processing using character data, and generates speech by combining phonemes or phoneme chains that correspond to the character data, so the speech time The length is not constant, and it is extremely difficult to edit a single sentence by controlling the device.

FIG. 1 shows a schematic configuration of an example of a conventional composite voice response device.

As shown in the figure, the word editing type voice response device basically consists of a main control device 1, a voice editing device 2, a word registration memory 3, a temporary storage memory 8, and a voice output device 4, which are configured into a composite type. In addition, a generation processing device 5vCv memory 6
and an arbitrary word generation device consisting of a buffer memory 7 is added.

When responding, the fixed sentence of the response sentence uses the word voice data in the word registration memory 3, and the variable sentence uses the word voice data generated by the arbitrary word generation device to obtain the desired voice output. be.

In the word editing type voice response device, word editing is performed as follows.

In other words, suppose the word length is 500m51 and the number of response lines is 16.
If it is a line, there are 5 words in the word registration memory 3 in advance.
It is divided into 00m5 and stored.

The main control device 1 generates a word number or word number to be sent to each line upon interruption from the audio editing device 2 every 500 m5, and sends it to the audio editing device 2.

In the audio editing device 2, the word number is temporarily stored in the temporary storage device 8, but this temporary storage device 8 has a two-sided structure.The word number from the main control device 1 is written on one side, and the word number is written on the other side. The word number written in the above interrupt is read out, and writing and reading are performed alternately every 500 m5.

This writing/reading is performed by the audio editing device 2, which reads the corresponding audio data from the word registration memory 3 according to the read word number of each line related to the above-mentioned interrupt, and then reads the corresponding audio data from the audio output device 4. The audio is edited by sending it to the .

Word editing is carried out as described above, but as is well known, in the case of arbitrary word generation devices, it is necessary to combine phonemes or phoneme chains to create word audio data, and it is simply a conventional word editing method. If only an additional connection is made to the voice response device, it is impossible to synchronize the two devices, and it becomes extremely difficult to control both devices and edit one sentence.

An object of the present invention is to realize a composite voice response device that can easily edit sentences by appropriately synchronizing a word editing type voice response device and an arbitrary word generation device.

For this purpose, the present invention starts the process of combining phonemes or phoneme chains in the arbitrary word generation device at the same clock as the cycle of the voice editing process, and stores the word voice parameters that are the processing results in a two-sided structure. (the word audio parameters are stored on one side of the storage device, and when the word audio parameters reach the audio parameters for one cycle, the processing is interrupted and the processing can be resumed at the clock of the next cycle), and from the other side of the storage device. is configured to read audio parameters, and the word editing device simplifies editing control by processing the reading of audio parameters as a processing result of the arbitrary word generation device in the same way as reading and editing the word audio parameters of the word editing device. It is.

Hereinafter, one embodiment of the present invention will be described in detail with reference to FIG.

FIG. 2 shows the configuration of a buffer memory section (corresponding to buffer memory 7 in FIG. 1), which is the main part of the present invention. A certain write counter 9a is
It operates as an address counter that is initialized by the write address signal S1 from the generation processing device 5 shown in the figure and incremented by the write signal S2, and the write address is sent to address selectors 9c and 9d.

The address selectors 9c and 9d are the audio editing device 2 of FIG.
They can operate alternately by the action of the 500m5 interrupt signal S5 and the inverter 9i, for example, the address selector 9c selectively outputs the write address from the write counter 9a to the buffer memory 9e. At this time, the address selector 9d selects and outputs the read address from the read counter 9b to the buffer memory 9f.

As a result, the audio data S8 from the generation processing device 5 is written into the buffer memory 9e, and the audio data at the time of the previous 500m5 interrupt signal is transferred from the buffer memory 9f to the audio data selector 9h.
This is what is output to.

The read counter 9b is initialized by the word number information S3 from the audio editing device 2 and is incremented by the read signal S4, but the address selectors 9c and 9d receive a 500m5 interrupt signal S5 from the audio editing device 2. Audio data S8 is written alternately every time, and a buffer memo 'J 9 e is written every time the signal is received.

From 9f onwards, the audio data written at the time of the 500m5 interrupt signal is alternately read out to the audio data selector 9h.

The audio data selector 9h selects audio data related to the generation processing device 5, that is, audio data from the buffer memories 9e and 9f, or audio data S6 from the word registration memory 3, from the audio editing device 2. Word number information S7
The audio data that has been selectively output is then provided to an audio output device.

Therefore, the composite voice response device of this embodiment operates as follows.

That is, the main control device 1 has a 500 m5 input from the audio editing device 2.
A word number corresponding to the voice sent to each line is generated by interruption, but in this case, if voice is to be sent from an arbitrary word generation device to a certain line, the word number data includes It is sent to the audio editing device 2 with a message attached thereto.

On the other hand, the line number and character sequence are sent to the arbitrary word generation device shown in Figure 2, and as a result, the arbitrary word generation device
(2) The phoneme chain data is read from the memory 6 and combined processing is performed, and the processing results are sent to the buffer memory 9.

As mentioned above, this buffer memory 9 has a two-sided structure, and audio data as a processing result is written on one side in response to a certain interrupt signal, and data is written from the other side in response to the above-mentioned interrupt signal. The selected audio data is read out to the audio data selector.

The writing of audio data to the buffer memory 9 is interrupted when 500m5 worth of audio data has been written, and the next 500m5 worth of audio data is written.
The write process is controlled to be restarted on the interrupt signal side.

In this manner, the main control device 1 sends a word number to the audio editing device 2 by a certain interrupt signal, and at the same time sends a character sequence to the generation processing device 5, and the word number is written on one side of the temporary storage memory 8. The generation processing device 5 is configured to generate voice data from a character sequence and write the voice data on one side of the buffer memory 9.

In this way, at the next interrupt signal, the audio editing device 2 reads the audio data from the word registration memory 3 or the buffer memory 9 based on the word number written at the time of the previous interrupt signal, and sends it to the audio output device 4. By sending the message, a voice response is made.

As explained above, in the present invention, the phoneme chain combination process in the arbitrary word generation device is started at the same cycle as the voice editing process, and the voice parameters that are the processing results are stored in the two-sided storage device. while ensuring that it is stored under certain conditions;
The word editing device controls the reading of voice parameters from the word registration memory or the above-mentioned two-sided storage device based on the word number from the main control device, and sends them to the voice output device. Therefore, it is an arbitrary word generation device. Even if , the voice editing processing section of the word editing type voice response device can easily edit the text without changing the processing method.

[Brief explanation of the drawing]

FIG. 1 is a schematic block diagram of an example of a conventional composite voice response device, and FIG. 2 is a block diagram of a two-sided buffer memory according to the present invention, which is especially provided for an arbitrary word generation device. 1...Main control device, 2...Audio editing device, 3...Word registration memory, 4...Audio output device, 5......・Generation processing device, 6...
... vC ■ memory, 7 ... buffer memory, 8
...Temporary storage memory, 9a...Write counter, 9b...Read counter, 9c, 9d. 9g...Buffer selector, 9e, 9f...
...Buffer memory, 9h...Audio data selector.

Claims (1)

[Claims] 1. Speech parameters are generated by combining phonemes or phoneme chains with a word editing type voice response device that divides speech into words and stores them in a memory, and sends out speech by editing the words read from the memory. A composite voice response device is arranged in parallel with an arbitrary word generation device, and configured to send fixed words by the word editing type voice response device and send variable words by the arbitrary word generation device, The arbitrary word generation device, which generates arbitrary words based on the character sequence sent from the main control device in response to an interrupt from the word editing type voice response device, performs a combination process of phoneme or phoneme chain data, and has a two-sided structure. The audio editing device stores a predetermined amount of time worth of audio data in the buffer memory of the device, and receives word numbers from the main control device by the interruption, and the audio editing device receives and receives word numbers from the main control device by the interruption, and the audio editing device stores the audio data for a predetermined period of time in the buffer memory of the device. 1. A voice response device, characterized in that the voice response device is configured to control a word registration memory and the buffer memory after identifying whether the interrupt is a message, and read and output voice data related to the interrupt to a voice output device from either one.