JP2007192931A - Voice pattern conversion/dubbing system, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a voice pattern conversion/dubbing system capable of performing dubbing accompanied by voice pattern conversion between different languages, and a program. <P>SOLUTION: A conversion filter generation section 102 of the voice pattern conversion/dubbing system 100 generates a voice pattern conversion filter 1 by learning voice data 3 for a filter uttered and collected in English by an actor 2 and voice data 5 for a filter uttered and collected in English by a voice actor 4. A voice pattern conversion section 103 converts the voice uttered in Japanese by the voice actor 4 to the voice of the actor 2 by using the voice pattern conversion filter 1. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a voice quality conversion dubbing system and a program for performing a dubbing accompanied by voice quality conversion.

Conventionally, there is known a voice quality conversion technique for converting a voice uttered by a certain speaker (former speaker) into a voice of another speaker (target speaker) (see, for example, Patent Document 1). In Patent Document 1, the voice of a person watching a drama is converted into the voice of an actor who plays a character in the drama. In Patent Document 1, a technique is used in which speech feature parameters are matched between the original speaker and the target speaker, and speech is synthesized from the speech feature parameters of the target speaker so that the time axis of the speech of the original speaker matches. Yes.
JP-A-4-240900

  When voices of actors appearing in overseas movies and dramas are converted into Japanese voices, Japanese voice actors are generally dubbed, but the voice quality of Japanese voice actors is There is also a demand from viewers to realize the dubbing with the voice quality of the actor who actually plays. In other words, famous foreign actors often have a unique voice quality as part of their character, but if the voice-over of the voice actor does not match the voice quality of the actor, It can make you feel uncomfortable or disappointing. In the conventional voice quality conversion technology described above, the voice of the former speaker who uttered the same dialogue as the target speaker is converted to the voice of the target speaker. It is necessary for foreign actors to speak Japanese lines. If an overseas actor speaks a Japanese line, even if the inflection of that line is unnatural, it will be replaced by an inflection of a Japanese voice actor, so the unnaturalness of the inflection is eliminated. However, in most cases, foreign actors do not speak Japanese, and even if they can speak, it is not practical to have all lines spoken in Japanese.

  The present invention has been made in order to solve the conventional problems as described above. A voice quality conversion dubbing system and a program capable of performing dubbing with voice quality conversion between different languages are provided. provide.

  In order to solve the above-described problem, based on the voices of the first speaker and the second speaker uttered in the first language, the voice of the second speaker is converted into the voice of the first speaker. Conversion filter creation means for creating the first speaker conversion filter, and voice uttered in a second language different from the first language by the second speaker using the first speaker conversion filter There is provided a voice quality conversion dubbing system comprising voice quality conversion means for converting the voice of the first speaker.

  According to the present invention, the second speaker can create the first speaker conversion filter by learning the voices of the first speaker and the second speaker uttered in the first language, so that the second speaker can It is possible to convert the voice uttered in step 1 to the voice of the first speaker using the first speaker conversion filter. Therefore, even if the first speaker cannot speak the second language, if the second speaker can speak the first language and the second language, the second speaker The voice uttered in the language can be easily converted into the voice of the first speaker. Therefore, it is possible to perform dubbing with voice quality conversion between different languages.

  According to a second aspect of the present invention, in the voice quality conversion dubbing system according to the first aspect, the conversion filter creating means applies the voices of the second speaker and the third speaker uttered in the second language. A second speaker conversion filter for converting the third speaker's voice into the second speaker's voice based on the second speaker conversion filter; Is used to convert the voice of the first speaker using the second speaker conversion filter, and the voice of the second speaker is converted to the voice of the first speaker using the first speaker conversion filter. It is characterized by being converted to speech.

  According to the present invention, even if the third speaker can speak only a second language different from the first language spoken by the first speaker, the first speaker can also speak the second language. The first speaker conversion filter was created by learning the voices of the first and second speakers spoken in the first language, even though they could not speak, and were spoken in the second language The second speaker conversion filter is created by learning the voices of the second speaker and the third speaker, so that the voice of the third speaker uttered in the second language is the voice of the first speaker. It is possible to convert to voice, and to perform voice-over with voice quality conversion between different languages.

According to a third aspect of the present invention, the voice conversion means uses the conversion filter obtained by synthesizing the second speaker conversion filter and the first speaker conversion filter, so that the third speaker is the second speaker. The voice uttered in the language is converted into the voice of the first speaker.
According to the present invention, by performing voice quality conversion using the synthesized conversion filter, it is possible to reduce the time for the conversion process compared to performing voice quality conversion using the second speaker conversion filter and the first speaker conversion filter. It can be shortened.

According to a fourth aspect of the present invention, in the voice quality conversion dubbing system according to any one of the first to third aspects, the conversion filter creation means is provided in a server device, and the voice quality conversion means is provided in a client device. It is characterized by.
According to the present invention, the client device can perform voice quality conversion using the conversion filter created by the server device.

  According to the fifth aspect of the present invention, the voice of the second speaker is converted to the voice of the first speaker based on the voices of the first speaker and the second speaker uttered in the first language. A conversion filter creating step for creating a first speaker conversion filter for converting into a first language, and a voice uttered in a second language different from the first language by the second speaker. There is provided a program for executing a voice quality conversion step of converting to the voice of the first speaker using a filter.

According to the present invention, it is possible to easily perform dubbing with voice quality conversion between different languages by storing the program in a computer.
According to a sixth aspect of the present invention, in the program according to the fifth aspect, in the computer, the voice of the second speaker and the third speaker uttered in the second language in the conversion filter creating step. Based on the processing for creating a second speaker conversion filter for converting the third speaker's voice into the second speaker's voice, and the voice quality conversion step, the third speaker The voice uttered in the language is converted to the voice of the second speaker using the second conversion filter, and the voice of the second speaker is converted to the first speaker using the first conversion filter. And a process of converting the sound into a voice.

  According to the present invention, the first conversion filter and the second conversion filter can be created by storing the program in a computer, and the third speaker can use the conversion filter to generate the second conversion filter. It is possible to convert the voice uttered in the language of the first speaker into the voice of the first speaker who speaks the first language, and to perform dubbing with voice quality conversion between different languages.

  According to the present invention, the second speaker can create the first speaker conversion filter by learning the voices of the first speaker and the second speaker uttered in the first language, so that the second speaker can It is possible to convert the voice uttered in step 1 to the voice of the first speaker using the first speaker conversion filter. Therefore, even if the first speaker cannot speak the second language, if the second speaker can speak the first language and the second language, the second speaker The voice uttered in the language can be easily converted into the voice of the first speaker. Therefore, it is possible to perform dubbing with voice quality conversion between different languages.

  In addition, even if the third speaker can speak only a second language different from the first language spoken by the first speaker, the first speaker and the second episode spoken in the first language The first speaker conversion filter is created by learning the voice of the speaker, and the second speaker conversion filter is created by learning the voices of the second and third speakers uttered in the second language. By doing so, it becomes possible to convert the voice uttered by the third speaker in the second language into the voice of the first speaker, and it is possible to perform dubbing with voice quality conversion between different languages. It becomes possible.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(First embodiment)
First, a first embodiment according to the present invention will be described. The voice quality conversion dubbing system 100 according to the first embodiment is composed of one or a plurality of devices.
In FIG. 1, the functional structure of the voice quality conversion dubbing system 100 is shown. As shown in the figure, the voice quality conversion dubbing system 100 includes a voice collection unit 101, a conversion filter creation unit 102, and a voice quality conversion unit 103.

The voice collecting unit 101 is configured to include a microphone and a recording device, and the voices uttered by the first speaker and the second speaker in the same language (first language) are collected by the microphone and stored in the recording device. Record as audio data.
The conversion filter creation unit 102 performs learning using the voice data of the first speaker and the second speaker collected by the voice collection unit 101, and converts the voice of the second speaker into the voice of the first speaker. Voice quality conversion filter 1 (first speaker conversion filter) is created. Here, the conversion filter includes a function for performing an operation, a conversion table, and the like. As a learning method, for example, a feature amount conversion method based on a known mixed normal distribution model (GMM) can be used. Any other known method can be used as the method for creating the filter 1.

The voice quality conversion unit 103 uses the voice quality conversion filter 1 created by the conversion filter creation unit 102 to convert the voice uttered by the second speaker in a language (second language) different from the first language into the first episode. To the voice of the other person.
The functions of the conversion filter creation unit 102 and the voice quality conversion unit 103 are realized by executing processing according to a program stored in a memory by a CPU included in each device constituting the voice quality conversion dubbing system 100.

Next, with reference to FIG. 2, an operation example of the voice quality conversion dubbing system 100 according to the present embodiment will be described. Here, the voice of actor 2 (first speaker) who can speak only English is set as the target voice, and the voice actor 4 (second speaker) of bilingual is used as the conversion source speaker to create Japanese dubbing voice data. This will be explained with an example.
First, the voice quality conversion filter 1 for converting the voice of the bilingual voice actor 4 into the voice of the actor 2 is created.

Specifically, about 50 English sentences for preparing the voice quality conversion filter 1 are prepared, and the actor 2 reads the English sentence, and the voice collecting unit 101 obtains the voice data 3 for filtering (step S101).
Here, according to a study by Carnegie Mellon University (http://www.speech.cs.cmu.edu/cgi-bin/cmudict), there are 39 types of English phonemes as shown in FIG. . On the other hand, there are 19 types of Japanese phonemes as shown in FIG. 5, and Japanese phonemes are associated with English phonemes as shown in FIG. The approximately 50 English sentences are designed to include all English phonemes associated with Japanese phonemes.

Next, the bilingual voice actor 4 who can speak English and Japanese fluently reads out the prepared English sentence, and the voice collecting unit 101 obtains the voice data 5 for filtering (step S102).
Next, the conversion filter creation unit 102 uses the feature conversion method based on the mixed normal distribution model (GMM) to convert the voice quality conversion filter 1 for converting the voice quality from the filter voice data 5 to the filter voice data 3. Create (step S103).

Hereafter, a feature conversion method based on a mixed normal distribution model (GMM) (for example, A. Kain and MWMacon, “Spectral voice conversion for text-to-speech synthesis,” Proc.ICASSP, pp.285-288, Seattle, USAMay) , 1998.) will be described in detail.
The feature amount x of the voice of the speaker as the conversion source and the feature amount y of the voice of the speaker as the conversion destination, which are associated with each frame in the time domain,

と表す。ここで、ｐは特徴量の次元数であり、Ｔは転置を示す。ＧＭＭでは、音声の特徴量ｘの確率分布ｐ（ｘ）を

It expresses. Here, p is the number of dimensions of the feature quantity, and T indicates transposition. In GMM, the probability distribution p (x) of the speech feature quantity x is

と表す。ここで、α_ｉはクラスｉの重み、ｍはクラス数である。また、Ｎ（ｘ；μ_ｉ，Σ_ｉ）はクラスｉでの平均ベクトルμ_ｉおよび共分散行列Σ_ｉを有する正規分布であり、

It expresses. Here, α _i is the weight of class i, and m is the number of classes. N (x; μ _i , Σ _i ) is a normal distribution having a mean vector μ _i and a covariance matrix Σ _i in class i,

と表される。次に、元話者の音声の特徴量ｘから目標話者の音声の特徴量ｙへと変換を行う変換関数Ｆ（ｘ）は、

It is expressed. Next, a conversion function F (x) for converting the feature amount x of the voice of the original speaker into the feature amount y of the target speaker's speech is:

と表される。ここで、μ_ｉ ^（ｘ）、μ_ｉ ^（ｙ）はそれぞれｘおよびｙのクラスｉでの平均ベクトルを表す。また、Σ_ｉ ^（ｘｘ）はｘのクラスｉでの共分散行列を示し、Σ_ｉ ^（ｙｘ）はｙとｘにおけるクラスｉでの相互共分散行列を示す。ｈ_ｉ（ｘ）は、

It is expressed. Here, μ _i ^(x) and μ _i ^(y) represent average vectors in class i of x and y, respectively. Σ _i ^(xx) represents a covariance matrix of x in class i, and Σ _i ^(yx) represents a mutual covariance matrix in class i of y and x. h _i (x) is

である。変換関数Ｆ（ｘ）の学習は、変換パラメータである（α_ｉ、μ_ｉ ^（ｘ）、μ_ｉ ^（ｙ）、Σ_ｉ ^（ｘｘ）、Σ_ｉ ^（ｙｘ））を推定することにより行われる。ｘおよびｙの結合特徴量ベクトルｚを

It is. Learning of the conversion function F (x) is performed by estimating conversion parameters (α _i , μ _i ^(x) , μ _i ^(y) , Σ _i ^(xx) , Σ _i ^(yx) ). The combined feature vector z of x and y is

と定義する。ｚの確率分布ｐ（ｚ）はＧＭＭにより

It is defined as The probability distribution p (z) of z is determined by GMM

と表される。ここで、ｚのクラスｉでの共分散行列Σ_ｉ ^（ｚ）および平均ベクトルμ_ｉ ^（ｚ）はそれぞれ

It is expressed. Here, the covariance matrix Σ _i ^(z) and the mean vector μ _i ^(z) in class i of ^z are respectively

と表される。変換パラメータ（α_ｉ、μ_ｉ ^（ｘ）、μ_ｉ ^（ｙ）、Σ_ｉ ^（ｘｘ）、Σ_ｉ ^（ｙｘ））の推定は、公知のＥＭアルゴリズムにより行うことができる。
学習にはテキストなどの言語情報は一切使用せず、特徴量の抽出やＧＭＭの学習はコンピュータを用いて全て自動で行う。

It is expressed. The conversion parameters (α _i , μ _i ^(x) , μ _i ^(y) , Σ _i ^(xx) , Σ _i ^(yx) ) can be estimated by a known EM algorithm.
Language information such as text is not used for learning, and feature extraction and GMM learning are all performed automatically using a computer.

Next, the speech obtained by translating the speech spoken by the actor 2 into Japanese is directly read out by the voice actor 4 in Japanese, and the voice data 6 dubbed into Japanese is obtained (step S104).
The voice quality conversion unit 103 converts the voice data 6 that has been dubbed through the voice quality conversion filter 1 to obtain the voice data 7 that has been dubbed (step S105). This dubbed voice data 7 is dubbed voice data in Japanese that resembles the voice of actor 2 who can speak only English.

  As described above, by creating the voice quality conversion filter 1 using the voice data of the actor 2 and the voice actor 4 in English, the voice of the voice actor 4 uttered in Japanese using the voice quality conversion filter 1 is the voice of the actor 2. It becomes possible to convert to. In this way, if the voice actor 4 who can speak both English and Japanese is prepared, even if the actor 2 cannot speak Japanese, the voice voiced by the voice actor 4 in Japanese will be the actor. It is possible to convert the sound into two. That is, it is possible to perform dubbing with voice quality conversion between different languages.

(Second Embodiment)
Next, a second embodiment according to the present invention will be described.
Similar to the voice quality conversion dubbing system 100 according to the first embodiment, the voice quality conversion dubbing system 100 according to the second embodiment includes a voice collection unit 101, a conversion filter creation unit 102, and a voice quality conversion unit 103. Yes.
The voice collection unit 101 according to the present embodiment includes the voices of the first and second speakers uttered in the first language and the second uttered in the second language different from the first language. The voices of the speaker and the third speaker are collected.

In addition to the function (the function of creating the voice quality conversion filter 1) included in the conversion filter creation unit 102 according to the first embodiment, the conversion filter creation unit 102 includes the second speaker uttered in the second language, Based on the voice of the third speaker, a voice quality conversion filter 8 (second speaker conversion filter) for converting the voice of the third speaker into the voice of the second speaker is created.
The voice quality conversion unit 103 converts the voice uttered by the third speaker in the second language into the voice of the second speaker using the voice quality conversion filter 8, and converts the voice of the second speaker into the voice quality conversion filter 1. Used to convert to the voice of the first speaker.

Next, with reference to FIG. 3, an operation example of the voice quality conversion dubbing system 100 according to the present embodiment will be described. Here, an example will be described in which dubbing voice data in Japanese is created with the voice color of actor 2 who can speak only English as the target voice color and voice actor 10 who can speak only Japanese as the conversion source speaker.
Up to the procedure for creating the voice quality conversion filter 1 (up to step S103) is the same as that of the first embodiment.

Next, in order to create a voice quality conversion filter 8 for converting the voice of a voice actor 10 who can speak only Japanese into the voice of a bilingual voice actor 4, about 50 Japanese sentences are prepared. The bilingual voice actor 4 reads out the Japanese sentence, and the voice collection unit 101 obtains the voice data 9 for filtering (step S201).
Next, the voice actor 10 who can speak only Japanese reads out the prepared Japanese sentence, and the voice collecting unit 101 obtains the voice data 11 for filtering (step S202).

Next, the conversion filter creation unit 102 uses a feature conversion method based on a mixed normal distribution model (GMM) to generate a voice quality conversion filter 8 for converting voice quality from the filter voice data 11 to the filter voice data 9. Create (step S203).
Next, the voice actor 10 reads out the speech obtained by translating the speech spoken by the actor 2 into Japanese, and obtains voice-dubbed audio data 12 (step S204).

The voice quality conversion unit 103 passes the dubbed voice data 12 through the voice quality conversion filter 8 and further passes through the voice quality conversion filter 1 to obtain the dubbed voice data 13 (step S205). This dubbed voice data 13 is dubbed voice data in Japanese that resembles the voice of actor 2 who can speak only English.
Thus, even if the voice actor 10 can only speak Japanese, the voice quality conversion dubbing system 100 creates the voice quality conversion filter 1 by learning the voices spoken by the actor 2 and the voice actor 4 in English. 4 and voice actor 10 each learns the voice uttered in Japanese and creates voice quality conversion filter 8, so that voice actor 10 speaks Japanese in voice using voice conversion filters 1 and 8. It becomes possible to convert to speech, and dubbing between different languages can be easily performed.

It is known that the voice quality conversion filter 8 and the voice quality conversion filter 1 can be synthesized to create a new voice quality conversion filter. The voice of the voice actor 10 is directly converted into the voice of the actor 2 by using the new voice quality conversion filter. Can be converted. By adopting such a conversion method, the conversion processing time can be halved.
Moreover, various aspects can be taken about the arrangement | positioning method to the apparatus of the voice quality conversion dubbing system 100 (the audio | voice collection part 101, the conversion filter production part 102, and the voice quality conversion part 103).

  For example, the voice quality conversion dubbing system 100 can be configured by a server device and a client device, the conversion filter creation unit 102 can be arranged in the server device, and the voice quality conversion unit 103 can be arranged in the client device. With this configuration, the voice quality conversion filters 1 and 8 created by the server device can be downloaded to the client device, and voice quality conversion can be performed by the client device.

In the first and second embodiments, Japanese dubbing voice data is created using one voice actor as the conversion source speaker. However, the present invention is not limited to this, and a plurality of voice actors may be used as the conversion source speakers. Good.
In this case, since there are a plurality of third speakers who are conversion source speakers for one second speaker, it is necessary to create a voice quality conversion filter 8 for each conversion source speaker. Using the filter voice data for each conversion source speaker collected by the voice collection unit 101, the conversion filter creation unit 102 creates a voice quality conversion filter 8 for each conversion source speaker. In this way, one voice actor 4 who can speak both English and Japanese is prepared, and by reading out the Japanese speech that the voice actor 10 performed, the voice actors can share each voice actor. It is possible to shorten the time required for the reading-out operation, and the work efficiency can be improved.

Hereinafter, an example of the operation of the voice quality conversion dubbing system 100 when a plurality of conversion source speakers are used will be described in detail with reference to FIGS. 1, 3, and 7. In this case, there are a plurality of third speakers (voice actors 10) shown in FIG. Here, FIG. 7 is a diagram illustrating a learning process when creating the conversion functions F and G and a voice quality conversion process using the conversion functions F and G.
When voice (voice sample data) is recorded from the recording target speaker, the conversion filter creation unit 102 sends the voice of the third speaker (referred to as the former speaker) and the voice of the second speaker (referred to as the intermediate speaker). Thus, the conversion function F (voice quality conversion filter 8) for converting the voice of the original speaker into the voice of the intermediate speaker is generated. Here, the voice of the original speaker and the voice of the intermediate speaker are recorded in advance by uttering the same approximately 50 sentences (one set of voice contents) of Japanese in the original speaker and the intermediate speaker. Use. Further, since there is one intermediate speaker (bilingual voice actor 4) and there are a plurality of original speakers, a conversion function F is created for each conversion source speaker. That is, learning is performed for each of a plurality of former speakers and one intermediate speaker. Therefore, the voice of one intermediate speaker is provided in common for each of a plurality of former speakers. As a learning method, for example, a feature amount conversion method based on a mixed normal distribution model (GMM) can be used as in the first and second embodiments. In addition to this, any known method can be used.

  Next, the conversion filter creation unit 102 creates a conversion function G (voice quality conversion filter 1) for converting the voice of the intermediate speaker into the voice of the first speaker (referred to as the target speaker). Here, the voice of the middle speaker and the voice of the target speaker are preliminarily spoken with about 50 sentences (one set of voice contents) of English for the intermediate speaker (bilingual voice actor 4) and the target speaker (actor 2). Use what you recorded and recorded.

Note that the format of the conversion functions F and G is not limited to a mathematical expression, and may be expressed in the form of a conversion table.
Next, the learning process of the conversion functions F and G and the conversion process using the conversion functions F and G will be described with reference to FIG. Here, there are two former speakers, and the sentences for learning speech of the conversion function F and the conversion function G are about 50 sentences (setA) in Japanese and about 50 sentences (setB) in English, respectively. Suppose there is.

  First, as shown in FIG. 7, the conversion filter creation unit 102 performs learning based on the voice setA of the original speaker (Src.1) and the voice setA of the intermediate speaker (In.), And performs a conversion function F ( Src.1 (A)) is created (step S301). Similarly, the conversion filter creation unit 102 performs learning based on the voice setA of the original speaker (Src.2) and the voice setA of the intermediate speaker (In.), And performs a conversion function F (Src.2 (A)). ) Is created (step S302).

Next, as shown in FIG. 7, the transform filter creation unit 102 is based on the speech of about 50 sentences (setB) in English of the intermediate speaker (In.) And the speech setB of the target speaker (Tag.1). Learning is performed to create a conversion function G1 (In. (B)) (step S303).
In the conversion process, as shown in FIG. 7, the voice quality conversion unit 103 converts any Japanese speech of the original speaker (Src.1) into an intermediate speaker using the conversion function F (Src.1 (A)). (In.) Voice is converted (step S304). Next, the voice quality conversion unit 103 converts the speech of the intermediate speaker (In.) Into the Japanese speech of the target speaker (Tag. 1) using the conversion function G1 (In. (B)) ( Step S305).

  Similarly, as shown in FIG. 7, the voice quality conversion unit 103 converts any Japanese speech of the original speaker (Src.2) into an intermediate speaker (In) using the conversion function F (Src.2 (A)). .) (Step S306). Next, the voice quality conversion unit 103 converts the speech of the intermediate speaker (In.) Into the Japanese speech of the target speaker (Tag. 1) using the conversion function G1 (In. (B)) ( Step S307).

  Note that a function of directly converting the voice of the original speaker into the voice of the target speaker may be provided using a function in which the conversion function F and the conversion function G are combined.

  Without being restricted by the type of language spoken by the target speaker, it can be used for dubbing with voice quality conversion between different languages.

It is a block diagram which shows the function structure of the voice quality conversion dubbing system which concerns on the 1st Embodiment of this invention. It is a figure for demonstrating the operation example of the voice quality conversion dubbing system which concerns on the same embodiment. It is a figure for demonstrating the operation example of the voice quality conversion dubbing system which concerns on the 2nd Embodiment of this invention. It is a figure which shows the kind of English phoneme. It is a figure which shows the kind of Japanese phoneme. It is a figure which shows the correspondence of a Japanese phoneme and an English phoneme. It is a figure which shows the learning process at the time of producing the conversion functions F and G, and the voice quality conversion process using the conversion functions F and G.

Explanation of symbols

100 Voice quality conversion dubbing system 101 Voice collection unit 102 Conversion filter creation unit 103 Voice quality conversion unit

Claims (6)

A first speaker conversion filter for converting the voice of the second speaker into the voice of the first speaker based on the voices of the first speaker and the second speaker uttered in the first language. A conversion filter creation means to create;
Voice quality conversion means for converting the voice uttered by the second speaker in a second language different from the first language into the voice of the first speaker using the first speaker conversion filter. Voice quality conversion dubbing system characterized by that. The conversion filter creating means includes
Second speaker conversion for converting the voice of the third speaker into the voice of the second speaker based on the voices of the second and third speakers uttered in the second language Create a filter
The voice quality conversion means includes
The voice of the third speaker uttered in the second language is converted to the voice of the second speaker using the second speaker conversion filter, and the voice of the second speaker is converted to the first speaker. The voice quality conversion dubbing system according to claim 1, wherein the voice conversion of the first speaker is converted using a conversion filter. The voice quality conversion means includes
Using the conversion filter obtained by synthesizing the second speaker conversion filter and the first speaker conversion filter, the voice of the first speaker is expressed as the voice uttered by the third speaker in the second language. The voice quality conversion dubbing system according to claim 2, wherein the voice quality conversion dubbing system is performed. The conversion filter creating means is provided in a server device,
The voice quality conversion dubbing system according to any one of claims 1 to 3, wherein the voice quality conversion means is provided in a client device. On the computer,
A first speaker conversion filter for converting the voice of the second speaker into the voice of the first speaker based on the voices of the first speaker and the second speaker uttered in the first language. A conversion filter creation step to be created;
A voice quality conversion step of converting a voice uttered by the second speaker in a second language different from the first language into a voice of the first speaker using the first speaker conversion filter; Program to let you. On the computer,
In the conversion filter creation step,
Second speaker conversion for converting the voice of the third speaker into the voice of the second speaker based on the voices of the second and third speakers uttered in the second language The process of creating a filter,
In the voice quality conversion step,
The voice uttered by the third speaker in the second language is converted to the voice of the second speaker using the second conversion filter, and the voice of the second speaker is converted to the first conversion filter. The program according to claim 5, further executing a process of converting to the voice of the first speaker using a voice.

Images