JPS60220652A - Speech synthesizing system of exchange - Google Patents

Abstract

PURPOSE:To make it possible that a receiving subscriber specifies and recognizes a transmitting subscriber when the receiving subscriber hears a translated synthesized speech from an exchange, by adding speech character information indicating individual characters of the transmitting subscriber to translated speech language information to synthesize a voice. CONSTITUTION:The speech from a transmitting subscriber 1 is subjected to A/D conversion by an A/D converter 5 in an exchange 2 and is translated automatically from a language A to a language B by a speech automatic translating device 9, and the translated code is synthesized in a speech synthesizing device 10 and is subjected to D/A conversion in a D/A converter 6 through a switching network 4 and is outputted as a speech signal and reaches a receiving subscriber 3. In this case, an operator of a switchboard 15 registers the sex and the age of the transmitting subscriber 1 in a transmitting subscriber individual character information register device 12 as voice individual character information parameters of the transmitting subscriber 1 in accordance with conversation with the transmitting subscriber 1. Simultaneously, a coefficient of partial self-correlation is obtained from voice information by a voice individual character information extracting device 11, and parameter information is added to this coefficient from the device 12, and they are synthesized by the device 10 and are transmitted to the receiving subscriber 3. The receiving subscriber 3 specifies and recognizes the speech of the translated language B as the speech from the transmitting subscriber 1 and receives it.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Section] The present invention relates to a speech synthesis method in an exchange, and particularly to a speech synthesis method in an exchange having an automatic speech translation function.

[Prior art]

Conventionally, in the exchange of this kind of food, the voice synthesis method of the voice information translated and converted from the meaningful language A to the language B is to use the exchange to synthesize a synthesized voice by editing specific voice information or voice segments that have been recorded in advance. A method is used in which the information is generated and transmitted to the other party. Therefore, there is a drawback in that the receiver on the other end cannot recognize the individuality of the speaker's voice, that is, always listens to a fixed translated synthesized voice without knowing who the speaker is.

[Purpose of the invention]

The purpose of the present invention is to eliminate the above-mentioned drawbacks by adding voice characteristic information indicating the voice personality of the sender to the translated spoken colloquial information and synthesizing the voice, so that the translated synthesized voice from the exchange is transmitted to the receiver of the other party. The purpose of the present invention is to provide a speech synthesis method in an exchange that allows a person to identify and recognize a person who is speaking when he or she hears the message.

[Structure of the invention]

According to the present invention, an exchange having an fP automatic translation device and a speech synthesis device coupled thereto, includes a voice personality information extraction device and a voice personality information registration device, and the voice translated by the voice automatic translation device A voice synthesis system in an exchange is obtained, which is characterized in that the speech personality information is synthesized by adding the voice personality extraction device and the speaker's voice characteristic information from the voice personality information registration device and transmitted to the other party.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a speech synthesis method in an exchange according to the present invention, and is composed of a transmitter 1, a receiver 3, and an exchange 2 that connects them. The exchange network 4 of the exchange 2 includes analog-to-digital conversion circuits 5 and 6 connected to the transmitter 1 and the receiver 3, respectively, a signal distribution trunk circuit 7, a synthetic voice output trunk circuit 8, and a switchboard connection circuit. A trunk circuit 14 is connected. A switchboard 15 is connected to the switchboard connection trunk circuit 14, and the switchboard 15 further connects to a data communication control bus 16.
It is connected to the exchange central processing unit (hereinafter referred to as CPU) 13 via. The automatic speech translation device 9 is a speech synthesis device 10
Both are connected to a signal distribution trunk circuit 7 and a synthetic voice output trunk circuit 8, respectively. Furthermore, the signal distribution trunk circuit 7 is also connected to the data reception buffer lla of the voice individuality information extraction device 11. The voice personality information extraction device 11 is connected to the voice synthesis device 10 via the data transmission buffer lid, and is connected to the speaker personality information registration device 1 via the central processing unit 11b.
It is also connected to 2. Here, audio & sexual information extraction device 1
1 includes a central processing unit flb, a data receiving buffer 11a1, a storage device 11 for storing a personality information extraction program;
C and a data transmission buffer lid. The speaker personality information registration device @12 is composed of a data input/output control circuit 12a and a storage device 12b connected thereto. It is connected to the data communication control bus 16.

Assuming that the speaker 1 is currently speaking to the receiver 3 in language A, the voice signal of the speaker 1 is encoded by the analog-to-digital conversion circuit 5 and sent via the normal line network path 4a. The signal is again decoded into an audio signal by the analog-to-digital conversion circuit 6 and transmitted to the receiver 3. Here, if the speaker 1 or receiver 83 wishes to translate the speaker 1's audio language A into language B, the CPU
13 releases the network path 4a and sets the network path 4b, so the caller is connected to the switchboard 15 via the switchboard connection trunk circuit 14. Exchange table 15
The operator inputs it into the speaker personality information registration device 12 as a parameter of the voice personality information of the speaker 1 and the speaker l. This input is set in the storage device 12b of the user individual information registration device 12 by a registration command prepared in the exchange 2.

Next, the exchange 2 releases the network path 4b.

The network path 4c and network path 4d are set and speech translation processing begins. That is, the speech signal of the speaker 1 encoded by the analog-to-digital conversion circuit 5 is input to the automatic speech translation device 9 via the network path 4c and the signal distribution trunk circuit 7;
The automatic speech translation device 9 recognizes and recognizes this input code sequence.
It is analyzed and translated into the language B, converted and generated into a code sequence corresponding to the language B, and the processing is handed over to the speech synthesis device 10. At the same time, the voice information of the speaker 1 sampled and encoded by the analog-to-digital conversion circuit 5 from the signal distribution trunk circuit 7 is also input to the voice individuality information extraction device 11 . The voice personality information extraction device 11 calculates a known partial autocorrelation coefficient based on this input voice information, and uses this to calculate the gender and age of the speaker 1 stored in advance in the storage device 12bK6. Parameter information indicating the degree is added and input as a characteristic parameter of the speaker 1 to the speech synthesis device 10 via the data transmission buffer lid.

This series of processing is performed by the central processing unit 11b by reading out the voice personality information extraction program stored in the memory device 11C for storing the personality information extraction program. The flow of this process will be explained using FIG. 2.

FIG. 2 is a flowchart showing an example of the voice individuality information extraction processing procedure in FIG. In the same figure, processing 1
Processing 3 to Processing 3 is a preliminary step for calculating the partial autocorrelation coefficient of Processing 4, and Processing 1 to Processing 4 are an example of the flow of conventionally known speech analysis feature extraction, and Processing 5 is according to the present invention. This is the processing of the added part. That is, following the voice analysis feature extraction process, process 5 (extraction of gender information and age information) is performed. The speech synthesis device 10 shown in FIG. 1 is a means for generating and synthesizing the pseudo speech of the speaker 1 based on this information.

In the exchange 2, the digital voice information generated and synthesized by the voice synthesizer 10 is taken out from the trunk circuit 8 for outputting synthesized voice and converted to analog/4G voice information via the network path 4d and the analog/digital conversion circuit 6, and then sent to the receiver. 3. Through these series of operations, the receiver 3 is able to specifically recognize and receive the voice translated from language A to language B and pseudo-synthesized by the exchange 2 as being the voice from the sender 1. .

〔Effect of the invention〕

As explained above, the speech synthesis method in the exchange of the present invention is such that a voice individuality information extraction device and a voice individuality information registration device are added to the exchange having an automatic speech translation device and a speech synthesis device. This has the effect of allowing the speaker to be identified and recognized as if he were speaking in the translated language.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of a speech synthesis method in an exchange according to the present invention, and FIG. 2 is a flowchart showing an example of the voice individuality information extraction processing procedure in FIG. - In the diagram, 1...Speaker, 2...
Exchange, 3...Receiver, 4...Exchange network, 4a to 4d...Network path, 5, 6...Analog-digital conversion circuit, 7
...Trunk circuit for signal distribution, 8...
Trunk circuit for synthetic speech output, 9... automatic speech translation device t, 1o old... speech synthesis device, 11...
...Voice personality information extraction device, 11a...Data reception bar...Data transmission buffer, 12...Speech personality information registration device t, 12a... -Data input/output control circuit, 12b exchange unit, 16...data communication control bus.

Claims (1)

[Claims] An exchange having a voiced automatic translation device and a speech synthesis device coupled thereto, including a voice personality information extraction device and a voice personality information extraction device, extracts the voice personality information from the voice language information translated by the voiced automatic translation device. A speech synthesis method in an exchange, characterized in that speech synthesis is performed by adding speech characteristic information of a speaker from a device and a speech personality information registration device, and transmits the synthesized speech to a receiver.