JP6631186B2 - Speech creation device, method and program, speech database creation device - Google Patents

Description

  The present invention relates to a speech creation device, a method, and a program for creating an accent information voice added to phoneme segment data, and a speech database creation device.

  In the creation of a speech database for speech synthesis, a portion for each segment section in units of phonemes is cut out from the speech data for learning to be speech element data, and the speech database is stored in the speech database using phoneme labels corresponding to the segments as keys. be registered. To improve the synthesis quality, accent information is added to the registered speech segment data. At this time, first, morphological analysis processing is performed on the text data corresponding to the voice data, and accent information is given to each phoneme obtained as a result. Next, for each segment, the phoneme of the segment is associated with the phoneme obtained by the morphological analysis processing, and accent information assigned to the phoneme is acquired, and registered in the speech database together with the phoneme piece data of the segment. Is done.

  Here, the learning speech data that is the basis of the phoneme segment data is not necessarily uttered with an accent based on the linguistic information obtained from the corresponding text data. Therefore, conventionally, there is known a technique of correcting accent information generated based on linguistic information obtained from text data based on fundamental frequency information obtained from speech of voice data (for example, Patent Documents 1 and 2). Technology).

JP 2012-189703 A JP-A-6-337691

  As described above, in order to add accent information to phoneme segment data, it is necessary to associate a phoneme for each segment obtained from speech data with a phoneme obtained by morphological analysis processing on text data corresponding to the speech data. Need to do.

  Here, breathing is performed in the actual utterance of the voice data, and the value of the voice data is silence or a value close thereto when the breathing is performed. Hereinafter, this timing is referred to as “blank timing”. When the audio data is divided into segments, the blank timing is detected as a segment having a phoneme indicating silence. On the other hand, this blank timing generally corresponds to a morpheme break, and punctuation is often present at this position. However, there are cases where punctuation is not detected. Further, when the blank timing is caused by stagnation, the blank timing may be, for example, the position of a phoneme in the middle of one morpheme on the linguistic information. In these cases, the correspondence between the phoneme for each segment and the phoneme obtained by the morphological analysis processing is not properly performed.

  However, in the above-described related art, since the occurrence of a blank timing is not considered, there is a problem that correct correspondence information may not be obtained for each phoneme piece data due to a shift in the correspondence. In addition, since the correspondence between the position of the accent information obtained from the linguistic information and the position of the fundamental frequency obtained from the speech of the voice data is also shifted, the correction of the accent information based on the fundamental frequency may not be successful. was there.

  SUMMARY OF THE INVENTION It is an object of the present invention to obtain correct accent information by considering blank timing of audio data.

In an example of the aspect, an acquisition process of acquiring first position data indicating at least one of an accent position and a delimiter position from input voice data, and morpheme data including a plurality of morphemes generated from text data corresponding to the voice data A second position data indicating at least one of a position of an accent given to the first character and a position of a break between a plurality of morphemes, and first position data obtained from the audio data; When the first and second position data do not match , the processing unit includes a processing unit that performs a process of adding the first position data to the morphological data instead of the second position data. In the processing, if the break position indicated by the first position data matches the break position indicated by the second position data, the second The position indicated by the position data, to impart information comma.

  According to the present invention, correct accent information can be obtained by considering the blank timing of audio data.

It is a block diagram of an embodiment of a voice information creation device according to the present invention. FIG. 4 is a diagram illustrating a data configuration example of segment data. It is a figure showing the example of segment data. It is a figure showing the example of data composition of morpheme data. It is a figure showing the example of morpheme data. It is a figure showing the example of data composition of morpheme label data. It is a figure showing the example of morpheme label data. It is a figure showing the example of data composition of label data for learning. FIG. 3 is a diagram illustrating a data configuration example of fundamental frequency data. FIG. 3 is a diagram illustrating an example of a hardware configuration of a computer that can realize the audio information creating device as software processing. It is a flowchart which shows the example of data provision processing. It is a flowchart (the 1) which shows the detailed example of a correspondence process. It is a flowchart (the 2) which shows the detailed example of a correspondence process. It is a figure showing the example of the morpheme data after adjustment. It is a flowchart which shows the detailed example of an accent correction process.

  Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings. In speech synthesis, a speech unit data string that best matches a synthesis target phoneme string is retrieved from a speech database, and the speech unit data is combined to synthesize speech data. The present embodiment relates to a technique for creating accent information to be added to phonemic piece data registered in voice data for voice synthesis for voice synthesis.

  Here, "speech synthesis" refers to a process of synthesizing voice data that, when text data representing a sentence is input, allows a computer to generate a voice similar to the voice when a human reads the text data. Say. The “phoneme piece data” refers to time-domain speech waveform data representing one phoneme. The “phoneme” refers to the smallest unit in the language that causes a difference in meaning. For example, when there is a word “any”, the words “a” and “r” are decomposed so that each has a difference in meaning. "A" "y" "u" "r" "u" form phonemes, respectively. In the speech synthesis processing, first, morphological analysis processing is performed on input text data while referring to a morphological dictionary to obtain morphological data. Here, the "morpheme" is the smallest unit in the language that takes its own meaning. For example, when there is text data that says "I have twisted all the realities toward myself." By the morphological analysis processing, the contents of the text data are, for example, “all”, “reality”, “to”, “all”, “self”, “no”, “ho”, “screw”, “ta”, “no”, “da” and Is broken down into 13 morphemes. As the next stage of the speech synthesis, the morpheme data is further decomposed to generate data (hereinafter, referred to as “morpheme label data” in the present embodiment) composed of a sequence of phonemes. For example, when there is one morpheme “all” obtained from the text data, the morphemes correspond to phonemes “a”, “r”, “a”, “y”, “u”, “r”, and “u”, respectively. Each morpheme label data having a label is generated. As the next stage of the speech synthesis, the above-mentioned speech database is searched for each morpheme label data for each morpheme label data, so that the phoneme piece data to which the same phoneme label as the phoneme label is added is searched. As the last stage of the speech synthesis, the speech data extracted for each morpheme label data are combined to synthesize speech data corresponding to the input text sentence.

  FIG. 1 shows an embodiment of a speech information creating apparatus 100 according to the present invention for creating accent information to be added to phoneme piece data registered in a speech database for speech synthesis. It is a block diagram of a form. The voice information creation device 100 includes a voice analysis unit 101, a language analysis unit 102, and a data addition unit 103, and finally outputs the learning label data 130. The voice analysis unit 101 includes a voice recognition unit 110, an acoustic model 111, and a fundamental frequency analysis unit 112. The language analysis unit 102 includes a morphological analysis unit 120 and a morphological dictionary 121.

  The speech recognition unit 110 executes speech recognition processing on the input speech data 113 in the learning time domain while referring to the acoustic model 111, so that each speech data 113 is continued by one phoneme. The segment is divided into segments, which are speech sections, and as a result, segment data 114 is output.

  FIG. 2 is a diagram illustrating a data configuration example of the segment data 114. As shown in this figure, segment data segment [0], segment [1],..., Segment [L-1] for each of a total of L phonemes obtained by the speech recognition processing by the speech recognition unit 110. Holds variable data of seg_id, phone, start, end, prev, and next, respectively. seg_id holds a segment ID (identifier). phone holds phoneme labels. "start" holds a start time (sample number). end holds the end time (sample number). prev holds a pointer to the immediately preceding segment data 114, and next holds a pointer to the immediately following segment data 114. If the current segment data 114 is, for example, segment [1], prev holds the start address of segment [0], and next holds the start address of segment [2]. If the current segment data 114 is, for example, the head data segment [0], prev holds a NULL value which is an undefined value. If the current segment data 114 is, for example, end data segment [L-1], next holds a NULL value. By connecting the segment data 114 using prev and next, even if the segment data 114 is added to, for example, the last address on the memory, the segment data 114 can be inserted at an arbitrary position.

  FIG. 3 is a diagram illustrating a specific example of the segment data. For example, in the segment data 114 of seg_id = 3, phone holds “a” as a phoneme label. “start” holds “730” (sample) as the start time. End holds “810” (sample) as the end time. prev holds “2” as seg_id of the previous segment data 114. “next” holds “4” as seg_id of the next segment data 114. In FIG. 3, the phone of each segment data 114 of the first seg_id = 0 or the seventeenth seg_id = 16 holds “#”. This “#” indicates that in the segment section corresponding to the segment data 114 in which they are registered, the value of the audio data 113 is determined to be zero or near zero, ie, silence.

  In particular, at the time of registration in a voice database (not shown), for each sample section of the segment indicated by the start and end of the segment data 114, the waveform data corresponding to the sample section is cut out from the voice data 113 to be speech unit data. You. Then, the phoneme piece data is registered in the speech database using a phoneme label representing a phoneme corresponding to the segment as a search key.

  At the time of speech synthesis, high-quality synthesized speech data cannot be obtained simply by searching the speech database for phoneme segment data corresponding to each phoneme label for each morpheme label data. Ideally, if speech segment data generated from speech data obtained when a human input text sentence is actually read aloud is registered in the speech database and searched, the best synthesized speech data can be obtained. However, since the storage capacity of the speech database is limited, it is not possible to register phoneme segment data corresponding to all text sentences. Thus, accent information is added to the phoneme segment data and registered in the speech database. On the other hand, at the time of speech synthesis, accent information is also generated at the time of morphological analysis processing on text data to be synthesized, and is added to each morpheme label data. Then, for each morpheme label data, when retrieving phoneme segment data corresponding to each phoneme label from the speech database, extracting phoneme segment data having accent information having a value close to the accent information given to the morpheme label data is extracted. To join. As a result, it is possible to select, from the speech database, the speech segment data which is pronounced in an accent state close to the text data to be speech-synthesized, thereby improving the quality of the synthesized speech data.

  Specifically, in the present embodiment shown in FIG. 1, the morphological analysis unit 120 constituting the language analysis unit 102 first inputs an utterance text 122 that is text data corresponding to the learning speech data 113, A morphological analysis process is performed on the utterance text 122 while referring to the morphological dictionary 121. As a result, the morphological analysis unit 120 outputs the morphological data 123 and the morphological label data 124. The morphological analysis unit 120 detects an accent position for each accent block determined based on the morpheme indicated by the morphological data 123. Here, the term “accent” refers to the relative height of the fundamental frequency of a phoneme or the relative strength of the signal strength of a phoneme in a word or word combination (range that can be expressed in one word), and the word or word combination is referred to as “accent”. Called "block." The “accent position” refers to the position of the mora whose fundamental frequency is relatively highest or the position of the mora whose signal strength is relatively highest in the accent block. “Mora” refers to a segmental unit of a sound having a certain time length in phonology, and is a convenient unit for expressing the position of an accent in an accent block. Next, for each morpheme label data 124, the morpheme analysis unit 120 adds, to the morpheme label data 124, accent information indicating the positional relationship with the accent position corresponding to the accent block to which the phoneme indicated by the data belongs. The “accent information” is, for example, the number of mora of the phoneme corresponding to the morpheme label data 124 from the accent position corresponding to the accent block to which the phoneme belongs to the mora to which the phoneme belongs (minus value before the accent position, plus Value) and a mora number indicating the number of the mora counted from the first mora in the accent block.

  FIG. 4 is a diagram illustrating a data configuration example of the morpheme data 123. As shown in this figure, each morpheme data morph [0], morph [1],..., Morph [M-1] for each of a total of M morphemes obtained by the morphological analysis processing by the morphological analysis unit 120. Holds variable data of morph_id, original, read, announcement, group, accent, prev, and next, respectively. morph_id holds a morpheme ID. original holds the morphological notation. read holds the notational reading. pronounce holds pronunciation. The group holds part of speech information. accent holds accent information. prev holds a pointer to the previous morpheme data 123 (the head is NULL). “next” holds a pointer to the next morpheme data 123 (the end is NULL). As in the case of the segment data 114, even if the morpheme data 123 is added to, for example, the end address on the memory by connecting the morpheme data 123 using prev and next, Can be inserted.

  FIG. 5 is a diagram showing a specific example of the morphological data 123. For example, when there is data of the utterance text 122 for learning having the content "All the reality has been twisted toward yourself." ”,“ All ”,“ self ”,“ no ”,“ ho ”,“ he ”,“ bend ”,“ ta ”,“ no ”,“ da ”, and“. ”. The notation, reading, The morpheme data 123 shown in FIG. 5 in which pronunciation is registered as original, read, and announcement is generated. For example, in the head morpheme data 123 with morph_id = 0, original holds “any” as a morpheme in notation. read holds "arayur" as a notational read. pronounce holds "arayur" as a pronunciation. The group holds “adnominal” as part of speech information. The Accent variable holds “3” as accent information. This indicates that there is an accent position at the third mora in this morpheme. Mora will be described later. prev holds “NULL” indicating that there is nothing before the morph_id of the morpheme data 123 immediately before. “next” holds “1” as the morph_id of the morpheme data 123 immediately after it.

  FIG. 6 is a diagram illustrating a data configuration example of the morpheme label data 124. As shown in this figure, a total of N morpheme label data morph_label [0], morph_label [1],..., Determined based on the morpheme data 123 obtained by the morphological analysis processing by the morphological analysis unit 120. morph_label [N-1] holds variable data of mlabel_id, phone, accent, mola, morph_id, prev, and next, respectively. mlabel_id holds a morpheme label ID. phone holds phoneme labels. accent holds the number of moras up to the accent position. mola holds the mora number. morph_id holds the morpheme ID (see FIG. 4) of the morpheme data 123 to which the phoneme indicated by the phoneme label held in the phone belongs. prev holds a pointer to the previous morpheme label data 124 (the head is NULL). “next” holds a pointer to the next morpheme label data 124 (the end is NULL). The meanings of prev and next are the same as in the case of the segment data 114 (FIG. 2) or the morphological data 123 (FIG. 4).

  FIG. 7 is a diagram illustrating a specific example of the morpheme label data 124. By the morphological analysis process, for example, each morpheme data having each morpheme ID of morph_id = 0,1,2,3 shown in FIG. 5 corresponding to four consecutive morphemes “all”, “reality”, “to”, and “all” From 123, 21 phonemes “a” “r” “a” “y” “u” “r” “u” “g” “e” “N” “j” “i” “ts” “u” “ "o", "s", "u", "b", "e", "t", and "e" are extracted, and 21 morpheme label data 124 in which phoneme labels corresponding to the respective phonemes are registered in the phone are generated. For convenience of processing, data having a phoneme label “#” indicating silence is registered as the morpheme label data 124 having a morpheme label ID of mlabel_id = 0.

  Here, for example, for the morphemes “all”, “reality”, “wo”, and “all” that have generated these morpheme label data 124 by the morphological analysis processing, the accent block “all” having the same delimiter as the delimiter of these morphemes "Reality", "to" and "all" are determined. Further, taking the accent block “any” as an example, four mora “a”, “ra”, “yu”, and “ru” are recognized in the accent block by the morphological analysis process. Then, by the morphological analysis processing, for example, the third mora “yu” (“yu”) is detected as an accent position in the accent block “any”. That is, in this example, the accent position is “3”.

  Further, by the morphological analysis processing, mlabel_id = 1,2,3,4,5 having each phoneme label corresponding to each of the phonemes “a” “r” “a” “y” “u” “r” “u” , 6, 7 in each morpheme label data 124, each mola variable contains the mora number “1” “2” “2” “3” “mora” of each mora including each phoneme in the accent block “any”. "3", "4" and "4" are retained. That is, in the morpheme label data 124 of mlabel_id = 1 and phone = “a”, the phoneme of the phoneme label “a” is the first of the four mora “a”, “ra”, “yu”, and “ru”. Is included in the mora “a”, so mola holds the mora number “1”. In the morpheme label data 124 of mlabel_id = 2 and phone = “r”, the phoneme of the phoneme label “r” is included in the second mora “ra” of the above four mora, so Holds the mora number “2”. Similarly, in the morpheme label data 124 of “a” of mlabel_id = 3 and phone =, the phoneme of the phoneme “a” is included in the second mora “ra” of the above four mora, so that mola Holds the mora number “2”. The same applies to other phonemes.

  In addition, by the morphological analysis process, mlabel_id = 1,2,3,4, having each phoneme label corresponding to each of the phonemes "a" "r" "a" "y" "u" "r" "u" In each of the morpheme label data 124 of 5, 6, and 7, each of the accent variables includes the mora number “1”, “2”, “2”, “3”, “3”, “4”, “4” And the difference value between the accent position "3" detected in the accent block "1-3 = -2", "2-3 = -1", "2-3 = -1", and "3-3 = 0, 3-3 = 0, 4-3 = 1, and 4-3 = 1 are registered.

  As described above, in FIG. 1, when the morpheme label data 124 in which the phoneme string including the accent information is registered by the morpheme analysis unit 120 in the language analysis unit 102 is obtained, the data adding unit 103 causes the speech analysis unit 101 The correspondence between the phoneme sequence of the segment data 114 generated by the voice recognition unit 110 and the phoneme sequence of the morpheme label data 124 is generated. For example, in the segment data 114 illustrated in FIG. 3 and the morpheme label data 124 illustrated in FIG. 7, the data adding unit 103 sequentially performs the phoneme label and the morpheme label of the phone variable of the segment data 114 from the head of each. It is checked whether the phoneme label of the phone variable in the data 124 matches. If the phoneme labels of both the segment data 114 and the morpheme label data 124 match, the data adding unit 103 adds the data structure of FIG. 6 to the contents of the segment data 114 having the data structure example of FIG. The accent information registered in the morpheme label data 124 having an example, that is, the number of mora up to the accent position and the mora number in the accent block are given, and output as the learning label data 130.

  FIG. 8 is a diagram illustrating a data configuration example of the label data for learning 130. The number of the learning label data 130 is basically the same as the L number of the segment data 114 illustrated in FIG. L learning label data train_label [i] (0 ≦ i ≦ L) corresponding to L segment data segment [i] (0 ≦ i ≦ L) are tlabel_id, phone, start, end, accent, Holds variable data of mola, pitch, vowel, prev, next. tlabel_id is a label ID for learning. The phone, start, and end are respectively copied from phone, start, and end (see FIG. 2) of the segment data segment [i]. The accent and mola are copied from the accent and mola (see FIG. 6) of the morpheme label data morph_label [j] (0 ≦ j ≦ N) associated with the segment data segment [i]. vowel is a flag indicating whether or not the phoneme label indicated by the phone is a vowel; "1" is set for a vowel and "0" is set for a consonant. prev holds a pointer to the immediately preceding learning label data 130, and next holds a pointer to the immediately succeeding learning label data 130. The pitch will be described later.

  Using the learning label data 130 thus obtained, the data of the section corresponding to the start and end (see FIG. 8) of the learning label data 130 is cut out from the voice data 113 of FIG. The phoneme segment data is registered in the speech database together with the phoneme label registered in the phone of the learning label data 130 in FIG. 8 and the accent information registered in the accent and mola. At the time of speech synthesis, an accent position is detected for each accent block even during morphological analysis of text data, and a value indicating a positional relationship with an accent position of an accent block to which the phoneme belongs is obtained for each phoneme in a phoneme string to be synthesized. Is done. When retrieving phoneme piece data corresponding to the phoneme label from the speech database, phoneme piece data to which accent information having a value closest to the value indicating the positional relationship obtained for each phoneme is added. Is extracted. As a result, it is possible to select phoneme piece data that is pronounced in the same accent state as the text sentence data, and it is possible to improve the quality of synthesized speech data.

  Here, as described in the section “Problems to be Solved by the Invention”, breathing occurs in the actual utterance of the audio data 113, and the timing is such that the value of the audio data 113 is close to a silence value or a silence value. It is a blank timing to take a value. When the voice data 113 is divided into segments, the blank timing is associated with the segment data 114 having the phoneme label “#” indicating silence. For example, the phoneme label “#” stored in the phone variable of each segment data 114 of the first seg_id = 0 or the 17th seg_id = 16 illustrated in FIG. 3 indicates that the segment corresponding to the segment data 114 is blank. Indicates a segment. On the other hand, this blank timing generally corresponds to a break of language information. In this case, when the morphological analysis process is performed, the blank timing is detected as a morpheme of punctuation. When the morpheme label data 124 is extracted from such punctuation morphemes, a phoneme label “#” indicating a blank timing is registered in the phone variable (see FIG. 6) of the morpheme label data 124. In such a state, when the data providing unit 103 associates the phoneme string of the segment data 114 with the phoneme string of the morpheme label data 124, the phoneme labels “#” match well, and the A correspondence is generated.

  On the other hand, in an actual utterance of voice data, punctuation may not be detected on linguistic information at the timing when silence timing occurs due to breathing. In this case, the phoneme label “#” indicating the blank timing is registered in the segment data 114, but the punctuation morpheme corresponding to the blank timing is not output, and the phoneme label “#” indicating the blank timing is stored in the phone variable. Are also not output. For example, in the example of the segment data 114 in FIG. 3, the breathing of the sound data 113 between the phoneme string “arayurugeNjitsuo” (notation: “every reality”) and the phoneme string “subete” (notation: “all”) is performed. Has occurred, seg_id = 16, phoneme label phone =, between segment data 114 of seg_id = 15, phoneme label phone = “o” and segment data 114 of seg_id = 17, phoneme label phone = “s”. The segment data 114 of the blank segment “#” has been generated. On the other hand, in the example of the morpheme data 123 of FIG. 5, no reading point is detected between the morpheme “o” and the morpheme “all”, and therefore, an example of the morpheme label data 124 of FIG. Also, between morpheme label data 124 of mlabel_id = 15, phoneme label phone = “o” and morpheme label data 124 of mlabel_id = 16, phoneme label phone = “s”, phoneme label phone = “#” The morpheme label data 124 of the blank segment has not been generated. In such a state, the data adding unit 103 determines whether the phoneme sequence of the segment data 114 illustrated in FIG. 3 and the phoneme sequence of the morpheme label data 124 illustrated in FIG. Are matched, the segment data 114 of seg_id = 15, phoneme label phone = “o” is matched with the morpheme label data 124 of mlabel_id = 15, phoneme label phone = “o”. Later, the segment data 114 of the blank segment with seg_id = 16 and phoneme label phone = “#” is compared with the morpheme label data 124 with mlabel_id = 17 and phoneme label phone = “s”. Since the phoneme labels do not match, the correspondence relationship is not established, and as a result, phoneme piece data based on the speech data 113 cannot be adopted for registration in the speech database.

  Therefore, in the present embodiment, the data adding unit 103 generates the correspondence between the phoneme string of the segment data 114 and the phoneme string of the morpheme label data 124 while determining the positional relationship between the blank segment and the morpheme break. Control is performed so that the correspondence does not shift. Details of this processing will be described later with reference to the flowcharts of FIGS.

  Here, the voice data 113 that is the basis of the phoneme segment data is not necessarily uttered at an accent position based on the morpheme data 123 obtained from the corresponding utterance text 122. Therefore, in the present embodiment, the data adding unit 103 determines the accent obtained from the morpheme label data 124 based on the correspondence between the phoneme sequence of the segment data 114 correctly generated as described above and the phoneme sequence of the morpheme label data 124. A new accent position for each block is recalculated as a position corresponding to a segment having the highest fundamental frequency extracted based on the speech of the audio data 113. Then, the data adding unit 103 corrects the accent information of the learning label data 130 corresponding to the segment belonging to the accent block based on the newly calculated accent position.

  More specifically, first, the fundamental frequency analysis unit 112 in the speech analysis unit 101 in FIG. 1 extracts the fundamental (pitch) frequency of the speech data 113 every predetermined frame period (for example, 256 milliseconds), 115 is output.

  FIG. 9 is a diagram illustrating a data configuration example of the fundamental frequency data 115. The K fundamental frequency data pitch [i] (0 ≦ i ≦ K−1) obtained as a result of analyzing the fundamental frequency of the audio data 113 for each predetermined frame period are time, pitch, prev, and next variables, respectively. Retain data. time holds the time corresponding to the current frame period (the sample number at the beginning, center, or end of the current frame period). The pitch holds the fundamental frequency [Hz] (Hertz) obtained as a result of the analysis. prev holds a pointer to the immediately preceding fundamental frequency data 115, and next holds a pointer to the immediately succeeding fundamental frequency data 115.

  Using the fundamental frequency data 115 generated by the fundamental frequency analysis unit 112 as described above, the data providing unit 103 generates a new accent position for each accent block, and generates a new accent position corresponding to a segment belonging to the accent block. The accent information of the label data 130 is corrected based on the newly calculated accent position. Details of this processing will be described later with reference to the flowchart in FIG.

  FIG. 10 is a diagram illustrating an example of a hardware configuration of a computer capable of realizing the functions of the voice analysis unit 101, the language analysis unit 102, and the data addition unit 103 of the voice information creation device 100 in FIG. 1 as software processing. In the computer shown in FIG. 10, a CPU 1001, a ROM (read only memory) 1002, a RAM (random access memory) 1003, an input device 1004, an output device 1005, an external storage device 1006, and a portable recording medium 1010 are inserted. And a communication interface 1008, which are interconnected by a bus 1009. The configuration shown in the figure is an example of a computer that can realize the above system, and such a computer is not limited to this configuration.

  The ROM 1002 includes a voice analysis processing program for realizing the function of the voice analysis unit 101 in FIG. 1, a language analysis processing program for realizing the function of the language analysis unit 102 in the figure, and data realizing the function of the data addition unit 103 in FIG. This is a memory for storing each program including the assignment processing program. The RAM 1003 is a memory that temporarily stores a program or data stored in the ROM 1002 when each program is executed.

  The external storage device 1006 is, for example, an SSD (solid state drive) storage device or a hard disk storage device, and includes the voice data 113, the uttered text 122, the segment data 114, the fundamental frequency data 115, the morpheme data 123, and the morpheme label data 124 in FIG. , Learning label data 130, and a voice database not shown in FIG.

  The CPU 1001 controls the entire computer by reading each program from the ROM 1002 to the RAM 1003 and executing the program.

  The input device 1004 detects an input operation by a user using a keyboard, a mouse, or the like, and notifies the CPU 1001 of the detection result.

  The output device 1005 outputs data transmitted under the control of the CPU 1001 to a display device or a printing device.

  The portable recording medium driving device 1007 accommodates a portable recording medium 1010 such as an optical disk, an SDRAM, and a compact flash, and has an auxiliary role of the external storage device 1006.

  The communication interface 1008 is a device for connecting, for example, a LAN (local area network) or WAN (wide area network) communication line.

  The system according to the present embodiment includes a voice analysis processing program for realizing the function of the voice analysis unit 101 in FIG. 1, a language analysis processing program for realizing the function of the language analysis unit 102 in the figure, and a function of the data providing unit 103 in FIG. Is read from the ROM 1002 to the RAM 1003 and executed by the CPU 1001. The program may be recorded on the external storage device 1006 or the portable recording medium 1010 and distributed, or may be obtained from the network by the network connection device 1008.

  The functions of the voice analysis unit 101 and the language analysis unit 102 in FIG. 1 realized by the CPU 1001 executing the voice analysis processing program and the language analysis processing program are as described above with reference to FIGS. 1 to 9. As a result, in the RAM 1003 of FIG. 10, the segment data 114 of the data configuration example of FIG. 2 (specific example is FIG. 3), the morphological data 123 of the data configuration example of FIG. 4 (specific example of FIG. 5), and the data configuration of FIG. The example morpheme label data 124 (a specific example is shown in FIG. 7) and the fundamental frequency data 115 of the data configuration example in FIG. 9 are generated.

  FIG. 11 is a flowchart illustrating an example of a data providing process for realizing the function of the data providing unit 103 in FIG. This data adding process is a process in which the CPU 1001 reads and executes a data adding process program stored in the ROM 1002 while using the RAM 1003 as a work memory.

  In the data adding process, the CPU 1001 first executes an associating process (step S1101).

  Next, the CPU 1001 determines whether or not an error has been detected as a result of the association processing (whether or not an error flag described later on the RAM 1003 is on) (step S1102). If no error is detected (NO in step S1102), CPU 1001 executes accent correction processing (step S1103). As a result, the CPU 1001 outputs the learning label data 130 (see FIG. 1) corresponding to the input voice data 113 and the utterance text 122, and ends the data adding process. If an error is detected (if the determination in step S1102 is YES), the CPU 1001 does not use the input voice data 113 and utterance text 122 for learning, and does not output the learning label data 130. Then, the data adding process is ended.

  FIGS. 12 and 13 are flowcharts showing a detailed example of the associating process in step S1101 of FIG. In this associating process, as described above, while the positional relationship between the blank segment and the morpheme break is determined, the phoneme string of the segment data 114 (see FIGS. 2 and 3) generated on the RAM 1003 is determined. An appropriate correspondence with the phoneme sequence of the morpheme label data 124 (FIGS. 6 and 7) is generated.

  In the association process, the CPU 1001 first determines whether the phoneme label on the segment data 114 side matches the phoneme label on the morpheme label data 124 side sequentially from the head of the set of the segment data 114 and the head of the set of the morpheme label data 124. (Step S1205), a process of determining whether the phoneme label on the side of the segment data 114 is "#" corresponding to a blank segment (step S1204), and a determination of step S1204 is YES. In the case of, the process of determining whether or not the phoneme label on the side of the morpheme label data 124 is “#” corresponding to the punctuation (Step S1208) is repeatedly executed.

  In order to control this repetition processing, the CPU 1001 sets the start address of the segment data 114 on the RAM 1003 and the start address of the morpheme label data 124 on the RAM 1003 in variables seg and morph on the RAM 1003, respectively (step S1201). In the example of the data configuration of the segment data 114 in FIG. 2, the address of segment [0] is set in the variable seg, and in the example of the data configuration of the morpheme label data 124 in FIG. 6, the address of morph_label [0] is set in the variable morph. You. Then, while it is determined that both the variable seg and the variable morph do not exceed the end data (when the determination in step S1202 is YES), the CPU 1001 repeatedly executes a series of processing from step S1203 to S1215 described below. .

  First, the CPU 1001 turns off both the increment flag on the side of the segment data 114 and the increment flag on the side of the morpheme label data 124 stored in the RAM 1003 (step S1203).

  Next, the CPU 1001 determines whether or not the phoneme label (the value of the phone variable in FIG. 2) on the side of the segment data 114 is “#” corresponding to a blank segment (step S1204).

  If the determination in step S1204 is NO, the CPU 1001 determines whether the phoneme labels of the phoneme label on the segment data 114 side and the phoneme label on the morpheme label data 124 side match (step S1205).

  The phoneme label on the segment data 114 side is not “#” indicating a blank segment (the determination in step S1204 is NO), and both the phoneme label on the segment data 114 side and the phoneme label on the morpheme label data 124 side are used. If the determination in step S1205 is NO because the phoneme labels do not match, it means that the voice data 113 and the utterance text 122 do not correspond well in the first place. Therefore, in such a case, the voice data 113 and the utterance text 122 input with an error should not be used for learning, and the learning label data 130 should not be output. Therefore, if the determination in step S1204 is NO and the determination in step S1205 is NO, the CPU 1001 sets the error flag stored in the RAM 1003 to ON (step S1207), and thereafter, FIGS. The association process of step S1101 in FIG. 11 shown in the flowchart of FIG. As a result, the determination in step S1102 of FIG. 11 is YES, and the accent correction processing in step S1103 is not executed, and the data labeling processing ends without generating the learning label data 130.

  If the determination in step S1205 is YES, the CPU 1001 turns on both the increment flag on the segment data 114 side and the increment flag on the morpheme label data 124 side in the RAM 1003 (step S1206).

  Thereafter, the CPU 1001 determines whether or not the increment flag on the side of the segment data 114 is ON (step S1212), and the determination in step S1212 is YES through the processing in step S1206. As a result, the CPU 1001 sets the address of the next segment data 114 in the variable seg (step S1213). Specifically, the CPU 1001 sets the value of the next variable (see FIG. 2) of the current segment data 114 to the variable seg.

  Subsequently, the CPU 1001 determines whether or not the increment flag on the side of the morphological data 123 is ON (step S1214), and the determination in step S1214 becomes YES through the processing in step S1206. As a result, the CPU 1001 sets the address of the next morpheme label data 124 in the variable morph (step S1215). Specifically, the CPU 1001 sets the value of the next variable (see FIG. 6) of the current morpheme label data 124 to the variable morph.

  After that, the CPU 1001 proceeds to the process of step S1202, and proceeds to the next repetitive process.

  As described above, when both the phoneme labels of the phoneme label on the side of the segment data 114 and the phoneme label on the side of the morpheme label data 124 match, both the variable seg and the variable morph are incremented, so that the next segment data The process proceeds to a process of associating the phoneme 114 with the phoneme of the next morpheme label data 124.

  If the determination in step S1204 is YES, the CPU 1001 determines whether the phoneme label (phone in FIG. 6) on the side of the morpheme label data 124 is “#” corresponding to punctuation (step S1208).

  If the determination in step S1208 is YES, the CPU 1001 sets both the increment flag on the segment data 114 side and the increment flag on the morpheme label data 124 side in the RAM 1003 to ON (step S1209).

  After that, the CPU 1001 proceeds to the above-described determination processing in step S1212, but the determination in step S1212 becomes YES through the processing in step S1209. As a result, the CPU 1001 proceeds to the process of step S1213 described above, and sets the address of the next segment data 114 in the variable seg.

  Subsequently, the CPU 1001 proceeds to the above-described determination processing in step S1214, and the determination in step S1214 becomes YES through the processing in step S1206. As a result, the CPU 1001 proceeds to the process of step S1215 described above, and sets the address of the next morphological data 123 in the variable morph.

  After that, the CPU 1001 proceeds to the process of step S1202, and proceeds to the next repetitive process.

  As described above, when the phoneme label on the side of the segment data 114 is “#” corresponding to the blank segment, and the phoneme label on the side of the morpheme label data 124 is “#” corresponding to the punctuation mark, and both correspond. Then, both the variable seg and the variable morph are incremented, and the process shifts to the process of associating the phoneme of the next segment data 114 with the phoneme of the next morpheme label data 124.

  If the determination in step S1208 is NO, the CPU 1001 inserts the current morpheme label in the morpheme data 123 obtained on the RAM 1003 in order to insert the morpheme label data of “@” before the current morpheme label data. After the morpheme data 123 including the phoneme of the label data 124, the morpheme data 123 indicating the reading point is inserted (step S1210). More specifically, the CPU 1001 first refers to the value of the morph_id variable (see FIG. 6) indicated by the variable prev of the current morpheme label data 124 on the RAM 1003 indicated by the variable morph, thereby obtaining the morpheme data 123 on the RAM 1003. The data that holds the same value as the value of the morph_id variable in the morph_id variable (FIG. 4) is searched. Next, the CPU 1001 generates an entry (see FIG. 4) of the new morpheme data 123 at the end of the morpheme data 123 on the RAM 1003. Then, the CPU 1001 assigns a new value to the morph_id variable and stores a value “,” representing a reading point in the original variable, the read variable, and the announcement variable in the newly generated entry at the end, and the group variable And the accent variable “0” is stored in the accent variable. Further, the CPU 1001 stores the value of the prev variable and the value of the morph_id variable of the retrieved morphological data 123 in the prev variable and the next variable in the entry newly generated at the end. Lastly, the CPU 1001 stores the value of the morph_id variable of the entry newly generated at the end in the next variable of the morphological data 123 indicated by the prev variable of the searched morphological data 123, and thereafter, the search is performed. The value of the prev variable of the morphological data 123 is also changed to the value of the morph_id variable of the entry newly generated at the end.

  For example, in the specific example of the segment data 114 in FIG. 3, the specific example of the morpheme data 123 in FIG. 5, and the specific example of the morpheme label data 124 in FIG. 7, when both the values of the variables seg and morph are “16”, In step S1204, it is determined that the phoneme label of the segment data 114 in which the value of the seg_id variable in FIG. 6 is “16” is “#”, and the determination result is YES. Further, in step S1208, the mlabel_id in FIG. Is determined to be “s” in the morpheme label data 124 whose value is “16”, and the determination result is NO. As a result, by executing step S1210, the CPU 1001 refers to the value “3” of the morph_id variable of the morpheme label data 124 whose mlabel_id value is “16” in FIG. The morpheme “all” whose value of the morph_id variable is “3” is searched. Next, the CPU 1001 generates an entry (see FIG. 4) of the new morpheme data 123 at the end of the morpheme data 123 of FIG. FIG. 14 shows a specific example of the morpheme data 123 after the entry is added. In the last entry of FIG. 14, the CPU 1001 assigns a new value “13” to the morph_id variable, stores the value “,” representing the reading point in the original variable, the read variable, and the announcement variable, respectively, and stores “ Symbol ”, and accent information“ 0 ”is stored in the accent variable. Further, the CPU 1001 assigns the value “2” of the prev variable and the value “3” of the morph_id variable of the entry of the morpheme “all” whose value of the morph_id variable retrieved on the morpheme data 123 of FIG. Are set in the prev variable and next variable of the last entry in FIG. Finally, the CPU 1001 determines the value of the prev variable “2” of the morpheme “all” entry whose value of the morph_id variable is “3” retrieved on the morphological data 123 of FIG. The value “13” of the morph_id variable of the last entry in FIG. 14 is stored in the next variable of the entry of the morpheme “wo” whose value is “2”, and the morph_id variable retrieved on the morpheme data 123 in FIG. The value of the prev variable of the entry of the morpheme “all” whose value is “3” is also changed to the value “13” of the morph_id variable of the last entry in FIG. As a result, morpheme data 123 shown in FIG. 14 is completed. As a result, after the morpheme “o” whose value of the morph_id variable is “2”, the reading value “13” of which the value of the morph_id variable is “13” is referred to by referring to the value “13” of the next variable of the entry. Is connected, and the value “3” of the next variable of the entry is referred to, thereby connecting the morpheme “all” entries of the value “3” of the morph_id variable.

  In the example described above, in the example of the segment data 114 in FIG. 3, between the phoneme string “arayurugeNjitsuo” (notation: “every reality”) and the phoneme string “subete” (notation: “all”), Due to the occurrence of breath in the pronunciation, seg_id = 16, phoneme between segment data 114 of seg_id = 15, phoneme label phone = “o” and segment data 114 of seg_id = 17, phoneme label phone = “s”. The segment data 114 of the blank segment with the label phone = “#” has been generated. On the other hand, in the example of the morpheme data 123 in FIG. 5, no reading point is detected between the morpheme “o” and the morpheme “all”. In such a case, in the present embodiment, the CPU 1001 can insert the reading point “,” between the morpheme “o” and the morpheme “all” on the morpheme data 123 in FIG. it can.

  After the processing of step S1210 in FIG. 12 described above, the CPU 1001 turns on the increment flag on the side of the segment data 114 in the RAM 1003 (step S1211). On the other hand, the increment flag on the side of the morpheme label data 124 on the RAM 1003 remains off in step S1203.

  After that, the CPU 1001 proceeds to the above-described determination processing in step S1212, but the determination in step S1212 becomes YES through the processing in step S1211. As a result, the CPU 1001 proceeds to the process of step S1213 described above, and sets the address of the next segment data 114 in the variable seg.

  Subsequently, the CPU 1001 proceeds to the above-described determination processing in step S1214, but the determination in step S1214 is NO due to the processing in step S1203. As a result, the CPU 1001 skips the processing in step S1215 described above and leaves the value of the variable morph at the current address of the morpheme data 123.

  After that, the CPU 1001 proceeds to the process of step S1202, and proceeds to the next repetitive process.

  As described above, the phoneme label on the side of the segment data 114 is “#” corresponding to the blank segment, and the phoneme label on the side of the morpheme label data 124 is not “#” corresponding to the punctuation mark. Is added, only the variable seg is incremented, and the process shifts to the process of associating the phoneme of the next segment data 114 with the phoneme of the current morpheme label data 124.

  In the specific example of the segment data 114 in FIG. 3, the specific example of the morphological data 123 in FIGS. 5 and 14, and the specific example of the morphological label data 124 in FIG. 7, when the values of the variables seg and morph are both “16”. Then, in step S1204, after it is determined that the phoneme label of the segment data 114 in which the value of the seg_id variable in FIG. 6 is “16” is “#” and the determination result is YES, the determination in step S1208 is NO. In step S1210, punctuation marks are inserted into the morpheme data 123 as shown in FIG. Thereafter, the processing proceeds from step S1211 → YES in S1212 to YES → S1213 → S1214 to NO → S1202, so that only the value of the variable seg is changed to the segment data 114 in which the value of the seg_id variable in FIG. Is incremented to the value “17” indicated by the next variable of As a result, the target of the next comparison processing is the phoneme label “s” of the segment data 114 in which the value of the seg_id variable in FIG. 3 is “17”, which is searched by the changed value “17” of the variable seg, and the variable The value of the mlabel_id variable in FIG. 7 is “16”, and the phoneme label “s” of the morpheme label data 124 is “16”.

  As a result of the above-described repetitive processing, if it is determined that either the variable seg or the variable morph has exceeded the end data (NO in step S1202), the CPU 1001 executes the processing in step S1216 and subsequent steps in FIG. I do. First, the CPU 1001 combines all notations of morphemes stored in the original variable of each morpheme data 123 shown in FIG. 14, for example, shown in FIG. The morphological analysis process (corresponding to the morphological analysis unit 120 in FIG. 1) is executed again, and the corresponding morphological data 123 and morphological label data 124 are generated on the RAM 1003 (step S1216).

  Thereafter, the CPU 1001 sequentially determines the phoneme label on the segment data 114 side and the phoneme label on the morpheme label data 124 side from the head of the set of the segment data 114 and the head of the set of the morpheme label data 124 newly generated on the RAM 1003. Are repeatedly determined (step S1220).

  Specifically, the CPU 1001 sets the start address of the segment data 114 on the RAM 1003 and the start address of the morpheme label data 124 newly generated on the RAM 1003 in the variables seg and morph on the RAM 1003 (step S1217). .

  Next, the CPU 1001 sets the address of the leading free area of the learning label data 130 on the RAM 1003 to the variable tr on the RAM 1003 to generate the learning label data 130 on the RAM 1003 (step S1218).

  While the CPU 1001 determines that both the variable seg and the variable morph do not exceed the end data (when the determination in step S1219 is YES), the CPU 1001 executes a series of processes from step S1220 to S1224 below. I do.

  First, the CPU 1001 determines whether or not both phoneme labels of the phoneme label on the segment data 114 side and the phoneme label on the morpheme label data 124 side match (step S1220).

  According to the processing of the flowchart of FIG. 12 described above, the morpheme data 123 of FIG. 14 is added to the segment data 114 of the blank timing generated by the occurrence of breathing or the like at the timing corresponding to the morpheme break in the utterance of the audio data 113. Correspondingly, additional readings have been inserted. Therefore, by performing the morphological analysis processing of step S1216 in FIG. 13 again on the text data obtained by recombining such morphological data 123, the phoneme sequence of the morpheme label data 124 generated again is This means that it is well associated with the phoneme string of the segment data 114.

  On the other hand, in a case where a blank timing occurs due to stagnation in the utterance of the audio data 113, the blank timing may be a position of a phoneme in the middle instead of a morpheme break. In this case, the segment data 114 at the blank timing is detected by the determination processing of step S1204 in FIG. 12, the determination result is YES, and the determination result of subsequent step S1208 is NO. The morpheme data 123 is forcibly inserted at the position of the morpheme break at the blank timing generated by the phoneme in the middle. In this state, when the morphological analysis process of step S1216 in FIG. 13 is executed again, the phoneme sequence of the morpheme label data 124 generated again does not correspond well to the phoneme sequence of the segment data 114. Originally, the speech data 113 containing the stagnation is not suitable for the production of the speech segment data. In such a case, the speech data 113 and the utterance text 122 input by generating an error are adopted for learning. Instead, the learning label data 130 should not be output.

  If the determination in step S1220 is NO because both the phoneme label on the segment data 114 and the phoneme label on the morpheme label data 124 do not match, the CPU 1001 stores the phoneme label in the RAM 1003. The error flag is set to ON (step S1224), and then the association processing of step S1101 in FIG. 11 shown in the flowcharts of FIGS. 12 and 13 is ended. As a result, the determination in step S1102 of FIG. 11 is YES, and the accent correction processing in step S1103 is not executed, and the data labeling processing ends without generating the learning label data 130.

  If the determination in step S1220 is YES because both the phoneme label of the segment data 114 and the phoneme label of the morpheme label data 124 match, the CPU 1001 returns to the RAM 1003 of the RAM 1003 indicated by the variable tr. The contents of the segment data 114 indicated by the variable seg and the contents of the morpheme label data 124 indicated by the variable morph are added to the new area of the learning label data 130 (see FIG. 8). Specifically, the CPU 1001 copies the contents of the phone, start, and end variables of the segment data 114 (see FIG. 2) indicated by the variable seg in the RAM 1003 indicated by the variable seg into the new area. The contents of each variable of “accent” and “mola” of the morpheme label data 124 (see FIG. 5) are copied. Also, in the new area, when the phoneme label of the phone variable indicates a vowel phoneme, the CPU 1001 sets the vowel variable (see FIG. 8) to a value “1” indicating a vowel to indicate the consonant phoneme. If it is, set the value "0" indicating consonant to the vowel variable. Further, the CPU 1001 sets the value of the new morpheme label ID in the tlabel_id variable in the new area. Further, in the new record area, the CPU 1001 sets the value of the morpheme label ID of the immediately preceding learning label data 130 (NULL value at the beginning) in the prev variable, and sets the next variable in the next variable. Is set to the value of the morpheme label ID of the learning label data 130 (a null value at the end). The content of the pitch variable is set in an accent correction process described later.

  After step S1221, the CPU 1001 sets the addresses of the next segment data 114 and the next morpheme label data 124 in the variables seg and morph on the RAM 1003, respectively (step S1222). Specifically, the CPU 1001 sets the value of the next variable (see FIG. 2) of the current segment data 114 and the value of the next variable (see FIG. 6) of the current morpheme label data 124 to the variables seg and morph, respectively. I do.

  Further, the CPU 1001 sets the address of the next free area of the learning label data 130 on the RAM 1003 to the variable tr on the RAM 1003 (step S1223).

  Thereafter, the CPU 1001 shifts to the processing of step S1219 in FIG. 13 and shifts to the next repetition processing.

  As a result of the above-described repetitive processing, when it is determined that either the variable seg or the variable morph has exceeded the end data (the determination in step S1219 is NO), the CPU 1001 is shown in the flowcharts of FIGS. The association processing in step S1101 in FIG. 11 ends. In this case, since the learning label data 130 is generated without any error, the determination in step S1102 in FIG. 11 is NO, and the accent correction processing in step S1103 in FIG. 11 is executed.

  FIG. 15 is a flowchart showing a detailed example of the accent correction processing in step S1103 of FIG. In the accent correction processing, the learning label data 130 generated in the RAM 1003 by the association processing of step S1101 (FIGS. 12 and 13) in FIG. 11 and the function of the fundamental frequency analysis unit 112 in the voice analysis unit 101 in FIG. A new accent position is generated for each accent block using the fundamental frequency data 115 generated in the RAM 1003 by the fundamental frequency analysis process corresponding to the, and the accent information of the learning label data 130 corresponding to the segment belonging to the accent block. Is corrected based on the above-described newly calculated accent position.

  First, the CPU 1001 calculates, for each of the learning label data 130 (see FIG. 8) in the RAM 1003, the average value of the fundamental frequency in the segment section determined by the value of the start variable and the value of the end variable of the learning label data 130. It is calculated (step S1501). Specifically, the CPU 1001 extracts, from the basic frequency data 115 (see FIG. 9) on the RAM 1003, each of the basic frequency data 115 in which the value of the time variable falls within the above-mentioned section, and extracts the pitch variable of the basic frequency data 115. Are extracted and the average value of the fundamental frequencies is calculated. Then, the CPU 1001 holds the calculated average value as the fundamental frequency of the segment section corresponding to the learning label data 130 in the pitch variable of the learning label data 130 (see FIG. 8).

  Next, the CPU 1001 sets the start address of the learning label data 130 generated on the RAM 1003 to the variable tr on the RAM 1003 in step S1502, and then sets the variable tr to the next address of the learning label data 130 in step S1507. Are sequentially set, and a series of processes from steps S1504 to S1506 are repeatedly executed until it is determined in step S1503 that the address of the variable tr has exceeded the end of the learning label data 130.

  In the repetition processing, the CPU 1001 first determines whether or not the segment corresponding to the learning label data 130 indicated by the variable tr is a break between accent blocks (step S1504). Specifically, the CPU 1001 determines that the value of the mora number indicated by the mola variable in the learning label data 130 indicated by the variable tr (see FIG. 8) is immediately before the value indicated by the prev variable in the learning label data 130. When the value of the mola number indicated by the mola variable in the learning label data 130 is smaller than the value of the mola variable, it is determined that the accent block to which the segment of the immediately preceding learning label data 130 belongs is finished, and the determination in step S1504 is made. Becomes YES. Alternatively, when the phone variable of the learning label data 130 indicated by the variable tr holds a phoneme label “#” indicating a blank segment, the CPU 1001 immediately precedes the preceding variable indicated by the prev variable in the learning label data 130. It is determined that the accent block to which the segment of the learning label data 130 belongs has ended, and the determination in step S1504 is YES.

  If the determination in step S1504 is YES, the CPU 1001 determines the segment of the segment determined by the value of the start variable and the value of the end variable (see FIG. 8) in the range of the accent block to which the segment of the immediately preceding learning label data 130 belongs. Of the learning label data 130 whose section belongs to the accent block, the phoneme position (for example, the value of the start variable) of the learning label data 130 having the highest fundamental frequency held by the pitch variable (see FIG. 8) is determined. Is determined as a new accent position (mora number). When the value of the vowel variable (see FIG. 8) of the learning label data 130 is “0”, that is, when the phoneme of the segment is a consonant, the phoneme positions of the learning label data 130 of the vowels before and after the consonant are used. Is set as a new accent position (step S1505).

  After step S1505, the CPU 1001 determines that the segment of the segment determined by the value of the start variable and the value of the end variable (see FIG. 8) in the range of the accent block includes the accent variable ( The difference between the mora number stored in the mola variable (see FIG. 8) and the new accent position calculated in step S1505 (the value before the accent position is a minus value, and the value after the accent position is a plus value). (Step S1506).

  After that, the CPU 1001 proceeds to the process in step S1507.

  If the determination in step S1504 is NO, the CPU 1001 skips the processing in steps S1505 and S1506 and proceeds to the processing in step S1507.

  In step S1507, the CPU 1001 sets the next address of the learning label data 130 to the variable tr. Specifically, the CPU 1001 stores the value of the next variable of the learning label data 130 (see FIG. 8) on the RAM 1003 indicated by the variable tr before the change into the variable tr. After that, the CPU 1001 returns to the process of step S1503.

  After the repetition processing, when it is determined that the address of the variable tr has passed beyond the end of the learning label data 130 and the determination in step S1503 is YES, the CPU 1001 proceeds to step S1103 in FIG. 11 shown in the flowchart of FIG. Is completed, and the learning label data 130 generated in the RAM 1003 is output to the external storage device 1006 or the like as the final learning label data 130 for creating a speech database of phoneme piece data.

  As described above, in the present embodiment, it is possible to obtain correct accent information by considering the blank timing of the audio data, and correct the accent information based on the fundamental frequency obtained from the audio data. It becomes possible.

  In the above-described embodiment, the accent information of each segment in the accent block is corrected according to the position of the segment having the highest fundamental frequency in the accent block. In addition, the accent information of each segment in the accent block may be corrected according to the position of the segment in which the signal strength is the maximum value in the accent block.

  The above-described embodiment is the embodiment of the accent information creating apparatus 100 for creating the learning label data 130 used for creating the speech database of the phoneme segment data. In addition, in addition to the voice analysis unit 101, the language analysis unit 102, and the data addition unit 103 shown in FIG. 1, a database registration unit is provided. A process of cutting out a portion to be performed as phoneme piece data, and registering the phoneme piece data and accent information that can be extracted from the learning label data 130 by the data providing unit 103 together with a phoneme label representing a phoneme of the segment in the speech database. The present invention may be embodied as an audio database creating device that executes processing.

Regarding the above embodiments, the following supplementary notes are further disclosed.
(Appendix 1)
An acquisition process of acquiring first position data indicating at least one of an accent position and a delimiter position from input voice data; and a process of obtaining first position data added to morpheme data including a plurality of morphemes generated from text data corresponding to the voice data. A second position data indicating at least one of a position of an accent and a break position between the plurality of morphemes, and a first position data obtained from the audio data; When the first and second position data do not match, the audio information includes a processing unit that executes a process of adding the first position data to the morphological data instead of the second position data. Creation device.
(Appendix 2)
The audio information creation device further includes:
2. The speech information creating device according to claim 1, further comprising a morphological analysis unit configured to generate morphological data including the plurality of morphemes by performing a morphological analysis process on the text data.
(Appendix 3)
3. The audio information creating apparatus according to claim 1, wherein the processing unit executes, as the acquisition process, a process of acquiring a position of a silent section of the audio data as a break position of the first position data.
(Appendix 4)
The processing unit may further include, in the comparison processing, when the break position indicated by the first position data matches the break position indicated by the second position data, the second position data is compared with the morphological data. 4. The audio information creation device according to any one of supplementary notes 1 to 3, wherein reading point information is added to the indicated position.
(Appendix 5)
The processing unit determines the fundamental frequency of the audio data in the section of the audio data corresponding to each of the plurality of morphemes as the acquisition processing. The audio information creation device according to any one of supplementary notes 1 to 4, wherein a process of acquiring a high position as an accent position of the first position data is performed.
(Appendix 6)
The processing unit determines the signal strength of the audio data in a section of the audio data corresponding to each of the plurality of morphemes as the acquisition processing. The audio information creation device according to any one of supplementary notes 1 to 4, wherein a process of acquiring a high position as an accent position of the first position data is performed.
(Appendix 7)
A voice information creation method used in a voice information creation device including a processing unit, wherein the processing unit,
Acquiring first position data indicating at least one of an accent position and a delimiter position from input voice data;
Second position data indicating at least one of a position of an accent given to morpheme data including a plurality of morphemes generated from text data corresponding to the speech data and a break position between the plurality of morphemes; Is compared with the first position data obtained from
If the first and second position data do not match, the audio information creating method includes adding the first position data to the morphological data instead of the second position data.
(Appendix 8)
In a computer used as a voice information creating device,
Acquiring first position data indicating at least one of an accent position and a delimiter position from input voice data;
Second position data indicating at least one of a position of an accent given to morpheme data including a plurality of morphemes generated from text data corresponding to the speech data and a break position between the plurality of morphemes; Comparing with the first position data obtained from
When the first and second position data do not match, adding the first position data to the morphological data instead of the second position data;
A program that executes
(Appendix 9)
An audio information creation device according to any one of supplementary notes 1 to 6,
Processing for cutting out speech element data from the speech data, and processing of the speech element data, phoneme labels representing the speech element data, and accent information given to the morpheme corresponding to the speech element data by the speech information creation device in a speech database. A registration processing unit for executing
An audio database creation device equipped with

REFERENCE SIGNS LIST 100 accent information creation device 101 voice analysis unit 102 language analysis unit 103 data addition unit 110 voice recognition unit 111 acoustic model 112 fundamental frequency analysis unit 120 morphological analysis unit 121 morphological dictionary 122 uttered text 1001 CPU
1002 ROM (read only memory)
1003 RAM (random access memory)
1004 Input device 1005 Output device 1006 External storage device 1007 Portable recording medium drive 1008 Communication interface 1009 Bus 1010 Portable recording medium

Claims (8)

An acquisition process of acquiring first position data indicating at least one of an accent position and a delimiter position from input voice data; and a process of obtaining first position data added to morpheme data including a plurality of morphemes generated from text data corresponding to the voice data. A second position data indicating at least one of a position of an accent and a break position between the plurality of morphemes, and a first position data obtained from the audio data; When the first and second position data do not match, a processing unit that performs a process of adding the first position data to the morphological data instead of the second position data ,
The processing unit may further include, in the comparison processing, when the break position indicated by the first position data matches the break position indicated by the second position data, the second position data is compared with the morphological data. An audio information creation device for adding reading point information to the indicated position . The audio information creation device further includes:
The voice information creation device according to claim 1, further comprising: a morphological analysis unit configured to generate morphological data including the plurality of morphemes by performing a morphological analysis process on the text data.   The audio information creation device according to claim 1, wherein the processing unit executes, as the acquisition process, a process of acquiring a position of a silent section of the audio data as a break position of the first position data. The processing unit determines the fundamental frequency of the audio data in the section of the audio data corresponding to each of the plurality of morphemes as the acquisition processing. the high position, executes a process of acquiring accent position of the first position data, sound information creating apparatus according to any one of claims 1 to 3. The processing unit determines the signal strength of the audio data in a section of the audio data corresponding to each of the plurality of morphemes as the acquisition processing. the high position, executes a process of acquiring accent position of the first position data, sound information creating apparatus according to any one of claims 1 to 3. A voice information creation method used in a voice information creation device including a processing unit, wherein the processing unit,
Acquiring first position data indicating at least one of an accent position and a delimiter position from input voice data;
Second position data indicating at least one of a position of an accent given to morpheme data including a plurality of morphemes generated from text data corresponding to the speech data and a break position between the plurality of morphemes; Is compared with the first position data obtained from
When said first and second position data do not match, the first position data assigned in place of the second position data to the morphological data,
When the delimiter position indicated by the first position data matches the delimiter position indicated by the second position data, reading point information is added to the position indicated by the second position data to the morphological data. , Audio information creation method. In a computer used as a voice information creating device,
Acquiring first position data indicating at least one of an accent position and a delimiter position from input voice data;
Second position data indicating at least one of a position of an accent given to morpheme data including a plurality of morphemes generated from text data corresponding to the speech data and a break position between the plurality of morphemes; Comparing with the first position data obtained from
When the first and second position data do not match, adding the first position data to the morphological data instead of the second position data;
When the delimiter position indicated by the first position data matches the delimiter position indicated by the second position data, reading point information is added to the position indicated by the second position data to the morphological data. Steps and
A program that executes An audio information creation device according to any one of claims 1 to 5 ,
A process of extracting phoneme data from the speech data, and processing the phoneme data, phoneme labels representing the phoneme data, and accent information given to the morpheme corresponding to the phoneme data by the speech information creating device in a speech database. And a registration processing unit for executing
An audio database creation device equipped with