JP5326546B2 - Speech synthesis dictionary construction device, speech synthesis dictionary construction method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To delete a mel-cepstrum portion series data having the possibility of constructing a voice synthesis dictionary bringing significant tone quality fluctuation, to prevent a synthesized voice from getting unnatural. <P>SOLUTION: A mel-cepstrum series data generating part 220 generates a mel-cepstrum series data from a voice data. A voice cut-out part 225 cuts out the mel-cepstrum portion series data in each phoneme from the mel-cepstrum series data. A mel-cepstrum data evaluating part 230 determines whether each mel-cepstrum portion series data reaches a voice feature amount numeric value having the possibility of constructing the voice synthesis dictionary bringing the significant tone quality fluctuation, or not. A mel-cepstrum data editing part 240 deletes the mel-cepstrum portion series data. A phoneme HMM learning part 250 finds thereafter a correspondence relation between a phoneme label and a phoneme mel-cepstrum data. A data writing-out part 260 writes the correspondence relation into the voice synthesis dictionary 202. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a speech synthesis dictionary construction device, a speech synthesis dictionary construction method, and a program for constructing a database used for speech synthesis.

  Speech recognition technology and speech synthesis technology based on a hidden Markov model (hereinafter referred to as HMM) are widely used as speech recognition and speech synthesis technology.

  Speech recognition technology and speech synthesis technology based on HMM are disclosed in Patent Documents 1 and 2, for example.

  In speech synthesis based on the HMM, a speech synthesis dictionary in which a correspondence relationship between phoneme labels and spectrum parameter data strings is recorded is required.

  A speech synthesis dictionary usually performs spectrum analysis and pitch extraction on data recorded in a database (hereinafter referred to as a speech database) composed of a set of phoneme label sequences and corresponding speech data. It is constructed through a learning process based on HMM.

  Conventionally, when constructing a speech synthesis dictionary, even if the mel cepstrum coefficients generated from the speech data contain significantly different mel cepstrum coefficients compared to other mel cepstrum coefficients, they are used as they are for learning based on the HMM. It was.

Japanese Patent Laid-Open No. 2002-244689 JP 2002-268660 A

  However, when speech synthesis is performed using the speech synthesis dictionary constructed as described above, an unnatural part may be included in the generated synthesized speech.

  For this reason, the synthesized speech using the speech synthesis dictionary constructed by the conventional speech synthesis dictionary construction device may be unnatural compared to the human natural speech.

  The present invention has been made in view of the above circumstances, and provides a speech synthesis dictionary construction device, a speech synthesis dictionary construction method, and a program capable of constructing a speech synthesis dictionary that does not synthesize unnatural speech. For the purpose.

In order to achieve the above object, a speech synthesis construction apparatus according to the first aspect of the present invention provides:
An apparatus for constructing a speech synthesis dictionary,
A receiver for receiving a phoneme label string and corresponding voice data;
A mel cepstrum sequence data generation unit for generating mel cepstrum sequence data from the entire audio data by generating a mel cepstrum coefficient in frame units from the received audio data;
A phoneme cutout unit that cuts out a mel cepstrum partial sequence data group corresponding to a phoneme label for each phoneme label from the generated mel cepstrum sequence data;
An average value is calculated for each extracted mel cepstrum partial series data group, and it is determined whether the mel cepstrum partial series data group includes mel cepstrum partial series data in units of frames whose difference from the average value exceeds a threshold value. If included, a mel cepstrum data evaluation unit that evaluates that the cut out mel cepstrum partial series data group is not suitable for constructing a speech synthesis dictionary;
A mel cepstrum data editing unit that edits the mel cepstrum partial series data by deleting a mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary;
A phoneme HMM learning unit that associates each phoneme label with a phoneme HMM related to the mel cepstrum by learning based on the hidden Markov model HMM from the phoneme label sequence and the edited mel cepstrum partial sequence data;
A data writing unit for recording the learning result in a speech synthesis dictionary;
It is characterized by providing.

The speech synthesis dictionary construction device according to the second aspect is
An apparatus for constructing a speech synthesis dictionary,
A receiver for receiving a phoneme label string and corresponding voice data;
A mel cepstrum sequence data generation unit for generating mel cepstrum sequence data from the entire audio data by generating a mel cepstrum coefficient in frame units from the received audio data;
A phoneme cutout unit that cuts out a mel cepstrum partial sequence data group corresponding to a phoneme label for each phoneme label from the generated mel cepstrum sequence data;
A first average value is calculated for each extracted mel cepstrum partial series data group, and second average values for all mel cepstrum partial series data groups belonging to the same phoneme label are calculated. A mel cepstrum data evaluation unit that evaluates that the mel cepstrum partial series data group having the first average value whose difference from the average value exceeds a threshold value is not suitable for constructing a speech synthesis dictionary;
A mel cepstrum data editing unit that edits the mel cepstrum partial series data by deleting a mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary;
A phoneme HMM learning unit that associates each phoneme label with a phoneme HMM related to the mel cepstrum by learning based on the hidden Markov model HMM from the phoneme label sequence and the edited mel cepstrum partial sequence data;
A data writing unit for recording the learning result in a speech synthesis dictionary;
It is characterized by providing.

  According to the present invention, speech synthesis is performed by deleting, from a mel cepstrum partial sequence data generated from speech data, features having significant feature amounts as a result of speech spectrum analysis compared to other mel cepstrum partial sequence data. Used for dictionary learning. For this reason, the synthesized speech obtained using the speech synthesis dictionary can be of high quality without causing an unnatural speech portion.

Hereinafter, the speech synthesis dictionary construction device according to the embodiment of the present invention will be described in detail.
(Embodiment 1)
First, the configuration of the speech synthesis dictionary construction device according to the present embodiment will be described.
FIG. 1 is a diagram illustrating a configuration of the speech synthesis dictionary construction device according to the first embodiment.

A speech synthesis dictionary construction apparatus 100 shown in FIG. 1 includes a ROM 101, a storage unit 102, a data input / output interface (I / F) 103, a user interface (I / F) 104, and a CPU 105. Are connected to each other.
The ROM 101 stores an operation program for learning based on the HMM, particularly an operation program including operations for evaluating and editing the mel cepstrum partial sequence data in this embodiment.

  The storage unit 102 includes a RAM 121 and a hard disk 122, and associates learning constants, phoneme label sequences, speech data, mel cepstrum sequence data, mel cepstrum partial sequence data, data phoneme labels, and phoneme mel cepstrum information. Remember things.

  The data input / output I / F 103 is an interface for connecting to the original data-containing hard disk 111 and the processed data recording hard disk 112 and the like.

  This data input / output I / F 103 is connected to the speech database 201 shown in FIG. 2, reads out a pair of phoneme label strings to be learned and speech data under the control of the CPU 105, and stores them in the storage unit 102.

  The data input / output I / F 103 is connected to the speech synthesis dictionary 202 shown in FIG. 2, and the correspondence between the phoneme label and the phoneme mel cepstrum information, which is the result of processing by the CPU 105, is shown in the speech synthesis dictionary shown in FIG. To 202.

  The user I / F 104 includes a keyboard (KB) 141 and a display (DSP) 142, and is provided for inputting arbitrary instructions, data, and programs. In particular, in the mel cepstrum data evaluation / editing process, the user needs to give a predetermined constant through the I / F.

  The CPU 105 executes the operation program stored in the ROM 101, thereby realizing each unit shown in the functional configuration of FIG.

FIG. 2 is a functional configuration diagram of the speech synthesis dictionary construction device 100 shown in FIG.
A speech synthesis dictionary construction apparatus 100 shown in FIG. 1 is connected to a speech database 201 and a speech synthesis dictionary 202 as shown in FIG.

  The voice database 201 is a database composed of a set of phoneme label strings and corresponding voice data, and is stored in a hard disk or the like.

  The speech synthesis dictionary 202 is a database constructed by the speech synthesis dictionary construction device 100 shown in FIG. 1, and stores phoneme labels and phoneme learning results in association with each other. This speech synthesis dictionary 202 is stored in a hard disk or the like.

  The phoneme learning result includes phoneme mel cepstrum information. Other pitch information necessary for speech synthesis varies depending on the specifications of the speech synthesizer, and the phoneme learning result includes such various information.

  As shown in FIG. 2, the speech synthesis dictionary construction apparatus 100 shown in FIG. 1 functionally includes a data extraction unit 210, a mel cepstrum sequence data generation unit 220, a phoneme segmentation unit 225, and a mel cepstrum data evaluation unit. 230, a mel cepstrum data editing unit 240, a phoneme HMM learning unit 250, and a data writing unit 260.

  The data extraction unit 210 reads data from the voice database 201 and separates it into a phoneme label string and voice data. The phoneme label string is delivered to the phoneme HMM learning unit 250, and the speech data is delivered to the mel cepstrum sequence data generation unit 220.

  The mel cepstrum sequence data generation unit 220 generates predetermined mel cepstrum sequence data from the audio data delivered from the data extraction unit 210 and delivers it to the phoneme segmentation unit 225.

  The phoneme cutout unit 225 cuts out the mel cepstrum partial sequence data for each phoneme from the mel cepstrum sequence data delivered from the mel cepstrum sequence data generation unit 220, and delivers it to the mel cepstrum data evaluation unit 230.

  The mel cepstrum data evaluation unit 230 evaluates whether each mel cepstrum partial series data delivered from the phoneme segmentation unit 225 is suitable for constructing a speech synthesis dictionary. Details of this evaluation process will be described later with reference to FIGS. 3 and 4.

  The mel cepstrum data editing unit 240 performs a predetermined editing process on the mel cepstrum partial sequence data delivered from the mel cepstrum data evaluation unit 230 to generate edited mel cepstrum partial sequence data. Details of the predetermined editing process will be described later with reference to FIGS.

  The edited mel cepstrum partial series data is delivered to the phoneme HMM learning unit 250.

  The phoneme HMM learning unit 250 converts the correspondence relationship between the phoneme label sequence and the edited mel cepstrum partial series data into a correspondence relationship (phoneme HMM) between the phoneme label and the phoneme mel cepstrum information by learning based on the HMM, and the correspondence relationship. Is delivered to the data writing unit 260.

  The data writing unit 260 records the correspondence between the phoneme label and the phoneme mel cepstrum information in the speech synthesis dictionary 202.

  The predetermined evaluation process executed by the mel cepstrum data evaluation unit 230 may be any process as long as it is a process for evaluating whether the mel cepstrum data is suitable for building a speech synthesis dictionary. Specific examples will be described.

  In the following description, a frame means a time segment used for generating a mel cepstrum coefficient and is represented by a symbol fm.

(Specific example 1 of evaluation / editing process)
A specific example 1 of the evaluation / editing process will be described with reference to the flowchart shown in FIG.

In this specific example, it is assumed that the user previously sets the threshold value th ^d mcep of the mel cepstrum coefficient difference in the storage unit 102 via the user I / F 104 of FIG.

  When the speech synthesis dictionary construction apparatus 100 shown in FIG. 1 constructs the speech synthesis dictionary 202, the speech synthesis dictionary construction apparatus 100 is connected to the speech database 201 and, for example, an empty speech synthesis dictionary 202. .

When an instruction to start speech synthesis dictionary 202 generated is from the user I / F 104, the data extraction unit 210, from the speech database 201, a phoneme label sequence MnLab _m and voice data Sp _m (however, 1 ≦ m ≦ _{M SP} Yes, _MSP is the number of data in the speech database.) The pairs are sequentially read out and stored in the storage unit 102.

Mel cepstrum series data generation unit 220, the voice data Sp _m from D following mel cepstrum series data Mcep ^d _m [fm] (where a 0 ≦ d ≦ D, 0 ≦ fm ≦ FLab m, FLab m audio data Is the number of frames for Sp _m .) And is stored in the storage unit 102 (step S1).

Then, the phoneme segmentation unit 225 performs phoneme segmentation on the mel cepstrum sequence data Mcep ^d _m using the phoneme label string MnLab _m to generate a mel cepstrum partial sequence data group mcep ^d _{L, n} (step S2). ). However, a _{0 ≦ L ≦ M L, M} L is a phoneme label total. Also, 0 ≦ n ≦ N _L , where N _L is the total number of mel cepstrum partial series data groups for the phoneme label.

The mel-cepstral data evaluation unit 230 determines that the mel-cepstrum partial series data group mcep ^d _{L, n} [fm] (where 0 ≦ fm ≦ F _{L, n} and FL _{, n} is the mel-cepstrum partial series data group mcep ^d _{L a} total number of frames in the _n.) for mel cepstrum partial series data group mcep ^d _L, the average value ave ^d _L for each _n, the _n is calculated according to the following equation (step S3).
ave ^d _{L, n} = {1 / (F _{L, n} + 1)} · Σ _{fm = 0} ^{FL, n} mcep ^d _{L, n} [fm]

Then, according to the following steps, for each phoneme label and for each mel cepstrum partial series data group mcep ^d _{L, n} , the frame unit mel cepstrum partial series data mcep ^d _{L, n} [fm] and the average value ave ^d _{L, n} The mel cepstrum partial series data group mcep ^d _{L, n} is evaluated based on the difference.

  First, a pointer L for designating a number for identifying a phoneme label is initialized to “1” (step S4).

Subsequently, a pointer n designating a number for identifying each mel cepstrum partial series data group is initialized to “1” (step S5).

Then, a pointer d for designating the dimension of each mel cepstrum partial series data is initialized to “0” (step S6).
With these pointers L, n, d, attention is focused on the mel cepstrum partial series data group mcep ^d _{L, n} [fm] (where 0 ≦ fm ≦ F _{L, n} ) on the storage unit 102.

  Then, the pointer fm indicating the frame number for the L, n-th and d-th order mel cepstrum partial sequence data is initialized to “0” (step S7).

Using this pointer fm, paying attention to the mel cepstrum partial sequence data mcep ^d _{L, n} [fm] of one frame, whether the difference from the average value for the frame fm to be processed is equal to or less than a predetermined threshold th ^d mcep The determination is made according to (1) (step S8).
| Mcep ^d _{L, n} [fm] −ave ^d _{L, n} | ≦ th ^d mcep (1)

If it is determined in step S8 that it is not less than the predetermined threshold value (step S8; No), the mel cepstrum partial series data group mcep ^d _{L, n} is deleted (step S9).

  If it is determined in step S8 that the value is equal to or less than the predetermined threshold (step S8; Yes), it is determined whether or not the processing has been completed for all fm (step S10).

  If it is determined that the process has not been completed (step S10; No), fm is incremented by “1” in step S14, and the process returns to step S8.

  If it is determined in step S10 that the processes for all fms have been completed (step S10; Yes), it is determined whether or not the processes have been completed for all d (step S11).

  If it is determined that the process has not been completed (step S11; No), d is incremented by “1” in step S15, and the process returns to step S7.

  If it is determined in step S11 that the processing for all d has been completed (step S11; Yes), it is determined whether or not the processing has been completed for all n (step S12).

  If it is determined that the process has not been completed (step S12; No), n is incremented by “1” in step S16, and the process returns to step S6.

  If it is determined in step S12 that the processing for all n has been completed (step S12; Yes), it is determined whether or not the processing has been completed for all L (step S13).

  If it is determined that the processing is not completed (step S13; No), L is incremented by “1” in step S17, and the process returns to step S5.

  As a result of determining whether or not the processing has been completed for all L in step S13, if it is determined that the processing has been completed (step S13; Yes), the processing ends.

Through the above processing, all mcep ^d _{L, n} [fm] that can be extracted from the speech database 201 of FIG. 2 and used for learning are stored in the storage unit 102 of FIG. The mcep ^d _{L, n} [fm] is used in the phoneme HMM learning unit 250 in FIG.

  In this specific example, the mel cepstrum partial sequence data used in the phoneme HMM learning unit 250 shown in FIG. 2 is mel cepstrum partial sequence data from which data that is significantly different from the mel cepstrum data editing unit 240 is removed in advance. This makes it possible to construct a speech synthesis dictionary that can synthesize more natural synthesized speech.

(Specific example 2 of evaluation / editing process)
A specific example 2 of the evaluation / editing process will be described with reference to the flowchart shown in FIG.

  In specific example 1, evaluation and editing are performed by repeating the process for the number of fm in step S8 in FIG. 3. However, if the fm value is large, a processing time is required.

  Therefore, in this specific example, instead of making a determination for each frame, the processing time is shortened by making a determination according to equation (2) described later.

  The flow of the operation of this example is as shown in FIG. 4 and is basically the same as that of Example 1 described with reference to FIG.

  Therefore, in FIG. 4, the same reference numerals are given to steps for performing the same processing as in FIG. 3.

  This example is different from Example 1 mainly in that Step S24 is provided, and Step S28 is provided instead of Step S8, and the loop of Steps S10 and S14 is removed.

In step S24, an average value AVE ^d _L of mel cepstrum partial series data belonging to the same monophone label is calculated according to the following equation.
AVE ^d _L = (1 / N _L ) · Σ _{n = 0} ^NL ave ^d _{L, n}

In step S28, the evaluation of the mel cepstrum partial series data for each monophone label is performed according to the following equation (2). The difference between the average values ave ^d _{L, n} and AVE ^d _{L of} each mel cepstrum partial series data group is This is done by determining whether a predetermined threshold is not exceeded.
| Ave ^d _{L, n} −AVE ^d _L | ≦ TH ^d mcep (2)

In this specific example, it is assumed that the user previously sets the threshold TH ^d mcep for the mel cepstrum coefficient difference in the storage unit 102 via the user I / F 104 in FIG.

  As described above, according to this example, a speech synthesis dictionary construction device that can output a natural synthesized speech without causing a sudden change in sound quality and can construct a speech synthesis dictionary efficiently. Can be provided.

  Two specific examples of the evaluation / editing process have been described above, but the evaluation / editing process is not limited to these. Any method can be used as long as it removes a significant difference in mel cepstrum data.

For example, for editing, instead of deleting the mel cepstrum partial series data that does not satisfy the conditions of the equations (1) and (2), the data mcep ^d _{L, n} before the data mcep ^d _{L, n} that does not satisfy the conditions _-1 or subsequent data mcep ^d _{L, n + 1} may be substituted.

  In the above, in order to facilitate understanding, an example in which all data is once read from the voice database 201 into the storage unit 102 by the data extraction unit 210 has been shown, but such batch processing is not an essential requirement of the present embodiment. For example, depending on the specifications of the phoneme HMM learning unit 250 shown in FIG. 2, it is possible to construct a speech synthesis dictionary more dynamically.

(Embodiment 2)
In the first embodiment, the phoneme label is associated with the phoneme mel cepstrum information by the speech synthesis dictionary construction device 100 shown in FIG. The present invention is not limited to this, and can also be applied to a case where a phoneme label is associated with phoneme mel cepstrum information and phoneme pitch information.

  The following describes the speech synthesis dictionary construction apparatus 500 that writes phoneme labels, phoneme mel cepstrum information, and phoneme pitch information in a speech synthesis dictionary in association with each other.

  As shown in FIG. 5, the speech synthesis dictionary construction apparatus 500 according to the present embodiment includes a data extraction unit 210 and a mel cepstrum data evaluation unit 230. These units have the same configurations and functions as the corresponding units of the speech synthesis dictionary construction apparatus 100 shown in FIG. 1 according to the first embodiment.

  The speech synthesis dictionary construction apparatus 500 further includes a sequence data generation unit 520, a phoneme segmentation unit 525, a data editing unit 540, a phoneme HMM learning unit 550, and a data writing unit 560.

The sequence data generation unit 520 generates pitch sequence data Pit _m by performing pitch extraction on the audio data extracted by the data extraction unit 210, in addition to the same function as the mel cepstrum sequence data generation unit 220. To do.

In addition to the same function as the phoneme segmentation unit 225, the phoneme segmentation unit 525 further performs a phoneme segmentation process on the pitch sequence data Pit _m to obtain the pitch subsequence data pit _{L, n} for each identical phoneme label. Generate.

In addition to the same function as the mel cepstrum data editing unit 240, the data editing unit 540 further evaluates that the mel cepstrum data evaluation unit 230 cannot be used as learning data and deletes the mel cepstrum partial series data mcep ^d _{L, n.} Pitch partial series data pit _{L, n} corresponding to is deleted.

  The phoneme HMM learning unit 550 converts the correspondence between the phoneme label string and the corresponding edited mel cepstrum sequence data into information indicating the correspondence between the phoneme label and the phoneme mel cepstrum information by learning based on the HMM. Then, the data is written to the data writing unit 560. Further, the phoneme HMM learning unit 550 learns the correspondence relationship between the phoneme label string and the pitch sequence data based on the HMM, converts it into information indicating the correspondence relationship between the phoneme label and the phoneme pitch information, and the data writing unit 560. To hand over.

  The data writing unit 560 writes the correspondence relationship between the phoneme label and the phoneme mel cepstrum information and the correspondence relationship between the phoneme label and the phoneme pitch information to the speech synthesis dictionary 202.

  By using the speech synthesis dictionary 202 constructed in this way, high-quality speech can be synthesized using phoneme mel cepstrum information and phoneme pitch information for each phoneme label.

(Embodiment 3)
In the third embodiment, the mel cepstrum partial sequence data mcep ^d _{L, n} is deleted in the first embodiment, and the mel cepstrum sequence data Mcep ^d _m obtained by cutting out the mel cepstrum partial sequence data is also deleted. is there.

  This is realized by making a change as shown in FIG. 6 in the specific example 1 of the evaluation / editing process shown in FIG. Hereinafter, the additional part (steps S61 to S68) will be described.

  After extracting the audio data (step S1), a counter m for counting the audio data is provided, an initial value “1” is set in the counter m (step S61), and whether or not the mel cepstrum partial series data has been deleted. And an initial value “0” is set in the flag DEL (step S63).

Then, after deleting the mel cepstrum partial series data mcep ^d _{L, n} (step S9), the deletion flag DEL is set to “1” (step S64).

For all the mel cepstrum partial series data indicated by the counter L, after the evaluation process in step S8 and the editing process in step S9 are completed (step S13; Yes), the deletion flag DEL is determined (step S65). when "1" is set, it deletes the mel cepstrum series data Mcep ^d _m (step S66).

  Thereafter, it is determined whether or not the processing has been completed for all audio data (step S67). If not, m is incremented (step S68), and the processing proceeds to the next audio data (step S63 and subsequent steps).

Further, in the above-described specific example 2 of the evaluation / editing process shown in FIG. 4, the mel cepstrum partial sequence data mcep ^d _{L, n} is deleted, and the mel cepstrum sequence data Mcep ^d _m obtained by cutting out the mel cepstrum partial sequence data is used. May also be deleted. In order to do this, a change as shown in FIG. 7 is made. The additional part of FIG. 7 is the same as the additional part of FIG.
Further, as a modification of the third embodiment, the mel cepstrum partial sequence data and the pitch partial sequence data are deleted, and the mel cepstrum partial data and the pitch partial sequence data extracted from the mel cepstrum partial sequence data and the pitch partial sequence data are also deleted. You may make it do. This is because the speech synthesis dictionary construction apparatus may generate again the mel cepstrum partial sequence data and pitch partial sequence data that are not suitable for construction of the speech synthesis dictionary from such mel cepstrum sequence data and pitch sequence data. It is. This change is realized by executing the processing of FIG. 6 or FIG. 7 by the apparatus of FIG.

(Embodiment 4)
In the fourth embodiment, the mel cepstrum partial sequence data and the mel cepstrum sequence data in the third embodiment are deleted, and further, the audio data generated from the mel cepstrum sequence data is also deleted. This is because such speech data may again generate mel cepstrum sequence data that is not suitable for construction of a speech synthesis dictionary. Then, the number of deleted audio data is counted, and when the counted value exceeds a predetermined threshold, all the audio data is recollected.

  This is realized by adding the changes shown in FIG. 8 to the above-described specific example 1 of the evaluation / editing process shown in FIG. Hereinafter, additional portions (steps S62, S80 to S83) will be described.

After taking out the audio data (step S1), a counter Cnt for counting the number of deleted audio data is provided, and an initial value “0” is set in the counter Cnt (step S62). The Mel cepstrum partial series data mcep ^d _L, to delete _n and mel-cepstrum series data Mcep ^d _m, deletes the voice data Sp _m that generated them (step S80), increments the counter Cnt (step S81 ).

  Then, after the processing is completed for all audio data (step S67; Yes), it is determined whether the count value of the counter Cnt has exceeded a predetermined threshold Thdel (step S82). When the count value of the counter Cnt exceeds a predetermined threshold Thdel (step S82; Yes), the CPU 105 in FIG. 1 notifies the user via the user I / F 104, for example, and re-records the user. And voice data is collected again (step S83).

Thereby, a higher quality speech synthesis dictionary can be constructed.
Further, in the above-described specific example 2 of the evaluation / editing process shown in FIG. 7, in order to delete the mel cepstrum partial series data and the mel cepstrum series data, and further delete the voice data generated from the mel cepstrum series data. A change as shown in FIG. 9 may be added. This is because there is a possibility that inappropriate partial series data may be cut out from these series data for the same reason as described at the end of the description of the third embodiment. The additional part of FIG. 9 is the same as the additional part of FIG.

Further, as a modification of the fourth embodiment, the mel cepstrum partial sequence data and pitch partial sequence data are deleted, and the mel cepstrum sequence data and pitch sequence data obtained by cutting out the mel cepstrum partial sequence data and pitch partial sequence data are deleted. Furthermore, the audio data that generated these data may also be deleted. This is because, for the same reason as described at the beginning of the description of the fourth embodiment, there is a possibility that inappropriate series data or even partial series data may be cut out from this audio data. This is realized by executing the processing of FIG. 8 or FIG. 9 by the apparatus of FIG.
In addition, this invention is not limited to the said embodiment, A various deformation | transformation and application are possible.

  For example, the above-described hardware configuration, block configuration, and flowchart are examples and are not limited.

  The present invention is not limited to the speech synthesis dictionary construction device, and can be constructed using any computer. For example, a program for causing a general-purpose computer to execute the above-described processing is distributed by a recording medium or communication, and this is installed and executed on the computer, so that it can function as the speech synthesis dictionary construction device of the present invention. Is possible.

It is a figure which shows the physical structure of the speech synthesis dictionary construction apparatus which concerns on Embodiment 1. FIG. It is a functional block diagram of the speech synthesis dictionary construction apparatus provided with the mel cepstrum data evaluation and edit part based on Embodiment 1. FIG. It is a figure which shows the flow of operation | movement in the specific example 1 of a mel cepstrum data evaluation and edit process. It is a figure which shows the flow of operation | movement in the specific example 2 of a mel cepstrum data evaluation and edit process. It is a function block diagram of the speech synthesis dictionary construction apparatus with pitch extraction based on Embodiment 2. FIG. It is a figure which shows the flow of operation | movement in the specific example 1 of a mel cepstrum data evaluation and edit process based on Embodiment 3. FIG. It is a figure which shows the flow of operation | movement in the specific example 2 of a mel cepstrum data evaluation and edit process based on Embodiment 3. FIG. It is a figure which shows the flow of operation | movement in the specific example 1 of a mel cepstrum data evaluation and edit process based on Embodiment 4. FIG. It is a figure which shows the flow of operation | movement in the specific example 2 of a mel cepstrum data evaluation and edit process based on Embodiment 4. FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Speech synthesis dictionary construction apparatus, 101 ... ROM, 102 ... Memory | storage part, 103 ... Data input / output I / F, 104 ... User I / F, 105 ... CPU, 106・ ・ ・ Bus, 111 ... Hard disk with original data, 112 ... Hard disk for recording processed data, 121 ... RAM, 122 ... Hard disk, 141 ... Keyboard, 142 ... Display, 201・ ・ ・ Speech database, 202 ・ ・ ・ Speech synthesis dictionary, 210 ・ ・ ・ Data extraction unit, 220 ・ ・ ・ Mer cepstrum sequence data generation unit, 225 ・ ・ ・ phoneme segmentation unit, 230 ・ ・ ・ Mer cepstrum data evaluation unit , 240... Mel cepstrum data editing unit, 250... Phoneme HMM learning unit, 260. ..Speech synthesis dictionary construction device, 520... Series data generation unit, 525... Phoneme segmentation unit, 540... Data editing unit, 550... Phoneme HMM learning unit, 560.

Claims (9)

An apparatus for constructing a speech synthesis dictionary,
A receiver for receiving a phoneme label string and corresponding voice data;
A mel cepstrum sequence data generation unit for generating mel cepstrum sequence data from the entire audio data by generating a mel cepstrum coefficient in frame units from the received audio data;
A phoneme cutout unit that cuts out a mel cepstrum partial sequence data group corresponding to a phoneme label for each phoneme label from the generated mel cepstrum sequence data;
An average value is calculated for each extracted mel cepstrum partial series data group, and it is determined whether the mel cepstrum partial series data group includes mel cepstrum partial series data in units of frames whose difference from the average value exceeds a threshold value. If included, a mel cepstrum data evaluation unit that evaluates that the cut out mel cepstrum partial series data group is not suitable for constructing a speech synthesis dictionary;
A mel cepstrum data editing unit that edits the mel cepstrum partial series data by deleting a mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary;
A phoneme HMM learning unit that associates each phoneme label with a phoneme HMM related to the mel cepstrum by learning based on the hidden Markov model HMM from the phoneme label sequence and the edited mel cepstrum partial sequence data;
A data writing unit for recording the learning result in a speech synthesis dictionary;
A speech synthesis dictionary construction device comprising: An apparatus for constructing a speech synthesis dictionary,
A receiver for receiving a phoneme label string and corresponding voice data;
A mel cepstrum sequence data generation unit for generating mel cepstrum sequence data from the entire audio data by generating a mel cepstrum coefficient in frame units from the received audio data;
A phoneme cutout unit that cuts out a mel cepstrum partial sequence data group corresponding to a phoneme label for each phoneme label from the generated mel cepstrum sequence data;
A first average value is calculated for each extracted mel cepstrum partial series data group, and second average values for all mel cepstrum partial series data groups belonging to the same phoneme label are calculated. A mel cepstrum data evaluation unit that evaluates that the mel cepstrum partial series data group having the first average value whose difference from the average value exceeds a threshold value is not suitable for constructing a speech synthesis dictionary;
A mel cepstrum data editing unit that edits the mel cepstrum partial series data by deleting a mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary;
A phoneme HMM learning unit that associates each phoneme label with a phoneme HMM related to the mel cepstrum by learning based on the hidden Markov model HMM from the phoneme label sequence and the edited mel cepstrum partial sequence data;
A data writing unit for recording the learning result in a speech synthesis dictionary;
A speech synthesis dictionary construction device comprising:   The mel cepstrum data editing unit deletes the mel cepstrum partial sequence data group evaluated as unsuitable instead of deleting the mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary. 3. The speech synthesis dictionary construction apparatus according to claim 1, wherein the mel cepstrum partial series data group belonging to the group is replaced. A method for building a speech synthesis dictionary,
A receiving step of receiving a phoneme label string and corresponding voice data from the database;
A mel cepstrum sequence data generation step for generating mel cepstrum sequence data from the entire audio data by generating a mel cepstrum coefficient in units of frames from the received audio data;
A phoneme segmentation step of segmenting a mel cepstrum partial sequence data group corresponding to a phoneme label for each phoneme label from the generated mel cepstrum sequence data;
An average value is calculated for each extracted mel cepstrum partial series data group, and it is determined whether the mel cepstrum partial series data group includes mel cepstrum partial series data in units of frames whose difference from the average value exceeds a threshold value. If included, a mel cepstrum data evaluation step for evaluating that the extracted mel cepstrum partial series data group is not suitable for constructing a speech synthesis dictionary;
A mel cepstrum data editing step for editing mel cepstrum partial series data by deleting a mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary;
A phoneme HMM learning step of associating each phoneme label with a phoneme HMM related to the mel cepstrum by learning based on a hidden Markov model HMM from the phoneme label sequence and the edited mel cepstrum partial sequence data;
An output step for outputting the learning result;
A speech synthesis dictionary construction method comprising: A method for building a speech synthesis dictionary,
A receiving step of receiving a phoneme label string and corresponding voice data from the database;
A mel cepstrum sequence data generation step for generating mel cepstrum sequence data from the entire audio data by generating a mel cepstrum coefficient in units of frames from the received audio data;
A phoneme segmentation step of segmenting a mel cepstrum partial sequence data group corresponding to a phoneme label for each phoneme label from the generated mel cepstrum sequence data;
A first average value is calculated for each extracted mel cepstrum partial series data group, and second average values for all mel cepstrum partial series data groups belonging to the same phoneme label are calculated. A mel cepstrum data evaluation step for evaluating a mel cepstrum partial sequence data group having a first average value whose difference from the average value exceeds a threshold value as being unsuitable for constructing a speech synthesis dictionary;
A mel cepstrum data editing step for editing mel cepstrum partial series data by deleting a mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary;
A phoneme HMM learning step of associating each phoneme label with a phoneme HMM related to the mel cepstrum by learning based on a hidden Markov model HMM from the phoneme label sequence and the edited mel cepstrum partial sequence data;
An output step for outputting the learning result;
A speech synthesis dictionary construction method comprising:   In the mel cepstrum data editing step, instead of deleting the mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary, the mel cepstrum partial series data group evaluated as inappropriate is used as the phoneme label. 6. The method of constructing a speech synthesis dictionary according to claim 4 or 5, wherein the mel cepstrum partial series data group belonging to is replaced. To a computer that executes the method of building a speech synthesis dictionary,
A receiving step of receiving a phoneme label string and corresponding voice data from the database;
A mel cepstrum sequence data generation step for generating mel cepstrum sequence data from the entire audio data by generating a mel cepstrum coefficient in units of frames from the received audio data;
A phoneme segmentation step of segmenting a mel cepstrum partial sequence data group corresponding to a phoneme label for each phoneme label from the generated mel cepstrum sequence data;
An average value is calculated for each extracted mel cepstrum partial series data group, and it is determined whether the mel cepstrum partial series data group includes mel cepstrum partial series data in units of frames whose difference from the average value exceeds a threshold value. If included, a mel cepstrum data evaluation step for evaluating that the extracted mel cepstrum partial series data group is not suitable for constructing a speech synthesis dictionary;
A mel cepstrum data editing step for editing mel cepstrum partial series data by deleting a mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary;
A phoneme HMM learning step of associating each phoneme label with a phoneme HMM related to the mel cepstrum by learning based on a hidden Markov model HMM from the phoneme label sequence and the edited mel cepstrum partial sequence data;
An output step for outputting the learning result;
A program that executes To a computer that executes the method of building a speech synthesis dictionary,
A receiving step of receiving a phoneme label string and corresponding voice data from the database;
A mel cepstrum sequence data generation step for generating mel cepstrum sequence data from the entire audio data by generating a mel cepstrum coefficient in units of frames from the received audio data;
A phoneme segmentation step of segmenting a mel cepstrum partial sequence data group corresponding to a phoneme label for each phoneme label from the generated mel cepstrum sequence data;
A first average value is calculated for each extracted mel cepstrum partial series data group, and second average values for all mel cepstrum partial series data groups belonging to the same phoneme label are calculated. A mel cepstrum data evaluation step for evaluating a mel cepstrum partial sequence data group having a first average value whose difference from the average value exceeds a threshold value as being unsuitable for constructing a speech synthesis dictionary;
A mel cepstrum data editing step for editing mel cepstrum partial series data by deleting a mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary;
A phoneme HMM learning step of associating each phoneme label with a phoneme HMM related to the mel cepstrum by learning based on a hidden Markov model HMM from the phoneme label sequence and the edited mel cepstrum partial sequence data;
An output step for outputting the learning result;
A program that executes   In the mel cepstrum data editing step, instead of deleting the mel cepstrum partial series data group evaluated as unsuitable for constructing a speech synthesis dictionary, the mel cepstrum partial series data group evaluated as inappropriate is used as the phoneme label. 9. The program according to claim 7, wherein the program is replaced with another mel cepstrum partial series data group belonging to.

Images