JP2005077663A - Voice synthesizer, voice synthesis method, and voice-synthesizing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enable BGM options suitable for a text to be read out, on the basis of a combinatorial conditions. <P>SOLUTION: A voice synthesizer extracts (S16) and detects (S14) environmental information affixed to an object text, and extracts category information (S22), sub-category information (S26), form information (S28), reading information (S32), ON/OFF information (S36), and partial information (S40) to be stored to a storing region of RAM. It generates a reading voice of the object text by voice synthesis processing (S42), and selects an appropriate BGM, on the basis of various conditions stored to RAM (S46) after calculating a time during reading (S44). When the texts is in series, common processing for selecting the same BGM is performed (S52), if a series of texts of the same category exists after repeating the above processing up to a final text. The generated reading voice is synchronized with the selected BGM and is output (S54). <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a speech synthesizer, a speech synthesis method, and a speech synthesis program that convert input text into speech and output, and more particularly, a speech synthesizer that outputs synthesized speech while reproducing BGM, and speech synthesis. The present invention relates to a method and a speech synthesis program.

When a text composed of a character string is read out by synthesized speech, an attempt is made not only to simply read out the character string but also to provide sound effects and BGM adapted to the text. For example, Patent Document 1 discloses a method of extracting a scene environment using an analysis result by natural language processing and outputting a sound effect corresponding to the environment. Patent Document 2 discloses a method of selecting an emotion to be given to a text to be read and reproducing background music in accordance with the emotion. Further, Patent Document 3 discloses a method of changing BGM playback conditions according to the content of information to be read out.
Japanese Unexamined Patent Publication No. 7-72888 JP 2001-166786 A JP 2001-34281 A

  However, in the above method, it is not possible to select a BGM that is adapted to the text in a complex manner based on information such as the contents of the text to be read out and the environment of the place where the text is read out.

  The present invention has been made to solve the above problem, and an object of the present invention is to be able to select a BGM suitable for text to be read out based on multiple conditions.

  In order to achieve the above object, a first invention of the present application is a speech synthesizer that reads out text by synthesized speech while reproducing BGM, and stores memory for storing BGM data classified at least by category and reproduction time. Means for extracting additional information added to the text, BGM selection means for selecting BGM data from the storage means based on the additional information extracted by the information extraction means, and the BGM selection means BGM reproducing means for reproducing the BGM data selected by the above in synchronization with the reading of the text.

  Further, according to the present invention, the additional information includes category information for classifying the content of the text, reading information for designating an attribute including a loudness, a pitch, and a tone for reading out the text, and the number of characters and the length of the text. May include either format information indicating the size or environment information indicating the volume of the surrounding sound to be read out.

  In the present invention, the BGM selection means may select BGM data based on at least one of the category information and the format information.

  In the present invention, the BGM selection means may select one piece of BGM data for each of the plurality of texts.

  In addition, the present invention includes a common unit that extracts a common feature from additional information of a plurality of continuous texts, and the BGM selection unit includes one unit for each of the plurality of texts shared by the common unit. BGM data may be selected.

  In the present invention, the common unit may extract the common category information as a common feature for a plurality of texts with the common category information.

  In the present invention, the additional information includes stop information for instructing whether or not to reproduce the BGM data, or partial information indicating a ratio of a portion for reproducing the BGM, and the stop information or the partial information Based on the above, BGM stop control means for starting and stopping the reproduction of the BGM may be provided.

  Next, according to a second aspect of the present application, there is provided an information extraction step of extracting additional information added to the text in a speech synthesis method in which text is read out by synthesized speech while reproducing BGM, and the information extraction step. A BGM selection step for selecting BGM data for the text from a plurality of BGM data classified and stored based on at least a category and a reproduction time based on the additional information extracted by the BGM data selected in the BGM selection step And a BGM reproduction step of reproducing in synchronization with the reading of the text.

  Further, according to the present invention, the additional information includes category information for classifying the content of the text, reading information for designating an attribute including a loudness, a pitch, and a tone for reading out the text, and the number of characters and the length of the text. May include either format information indicating the size or environment information indicating the volume of the surrounding sound to be read out.

  In the BGM selecting step, BGM data may be selected based on at least one of the category information and the format information.

  In the BGM selecting step, one BGM data may be selected for each of the plurality of texts.

  In addition, the present invention includes a common process for extracting common features from additional information of a plurality of continuous texts. In the BGM selection process, one text is used for a plurality of texts shared in the common process. BGM data may be selected.

  Moreover, this invention may extract the said common category information as a common characteristic about the some text in which the said category information is common in the said commonization process.

  In the present invention, the additional information includes stop information for instructing whether or not to reproduce the BGM data, or partial information indicating a ratio of a portion for reproducing the BGM, and the stop information or the partial information A BGM stop control step for starting and stopping the reproduction of the BGM may be provided based on the above.

  Next, the third invention of the present application causes a computer to execute the second invention.

  In the first invention of the present application, the BGM is selected based on the additional information added in advance to the text, so that the text can be read out while reproducing the BGM suitable for the text.

  Further, according to the present invention, since BGM can be selected according to a plurality of additional information, it is possible to listen to text reading while reproducing BGM more suitable for text.

  Further, according to the present invention, since BGM is selected based on category information or format information, it is possible to listen to text-to-speech while playing back a BGM that matches the genre of the text or a BGM that matches the length of the text. .

  Further, according to the present invention, since one BGM is selected for each text, the text can be read out while reproducing the BGM most suitable for each text.

  Further, according to the present invention, when a plurality of continuous texts have a common feature such as a common category, one BGM is selected for the common text, so that the BGM is frequently switched. And can read aloud text while playing back an appropriate BGM.

  Further, according to the present invention, when a plurality of continuous text categories are common, one BGM is selected for the common text, so that the BGM is not frequently switched and is appropriate. You can listen to text reading while playing BGM.

  Furthermore, since the present invention can instruct the start / stop of BGM playback as additional information, it is possible to listen to the text being read out while playing back the BGM only in the portion of the text where the BGM is to be played back.

  Next, in the second invention of the present application, since BGM is selected based on additional information added in advance to the text, it is possible to listen to the text while reading the BGM suitable for the text.

  Further, according to the present invention, since BGM can be selected according to a plurality of additional information, it is possible to listen to text reading while reproducing BGM more suitable for text.

  Further, according to the present invention, since BGM is selected based on category information or format information, it is possible to listen to text-to-speech while playing back a BGM that matches the genre of the text or a BGM that matches the length of the text. .

  Further, according to the present invention, since one BGM is selected for each text, the text can be read out while reproducing the BGM most suitable for each text.

  Further, according to the present invention, when a plurality of continuous texts have a common feature such as a common category, one BGM is selected for the common text, so that the BGM is frequently switched. And can read aloud text while playing back an appropriate BGM.

  Further, according to the present invention, when a plurality of continuous text categories are common, one BGM is selected for the common text, so that the BGM is not frequently switched and is appropriate. You can listen to text reading while playing BGM.

  Furthermore, since the present invention can instruct the start / stop of BGM playback as additional information, it is possible to listen to the text being read out while playing back the BGM only in the portion of the text where the BGM is to be played back.

  Next, the 3rd invention of this application can have the effect of the 2nd invention.

Next, the best mode for carrying out the present invention will be described with reference to the drawings.
FIG. 1 is a circuit block diagram of a portable information terminal 1 showing an example of a speech synthesizer according to the present invention. The portable information terminal 1 controls a CPU 10 that controls the entire portable information terminal 1, a memory control unit 20 that controls memories such as a nonvolatile memory 21 that stores various programs and databases, a RAM 22 that stores various data, and peripheral devices. A peripheral control unit 30 is connected. Connected to the peripheral control unit 30 are a microphone 31 that picks up ambient sounds and the like, a display 32, an input unit 33, an audio unit 34, and a speaker 35 that outputs sound sent from the audio unit 34. . Here, the audio unit 34 converts the audio synthesized by the CPU 10 into an analog audio signal, performs predetermined amplification, and sends the amplified audio signal to the speaker 35.

  Next, the nonvolatile memory 21 will be described with reference to FIGS. FIG. 2 is a block diagram schematically showing the configuration of the nonvolatile memory 21 that is a storage means. FIG. 3 is a block diagram schematically showing the configuration of the BGM database 211. 4 and 5 are block diagrams schematically showing the configuration of the reading information database 212. As shown in FIG. As shown in FIG. 2, the non-volatile memory 21 includes a BGM database 211 used for BGM selection, a reading information database 212 that stores information related to text reading, and a text that is used for speech synthesis. A speech synthesis dictionary 215 and a syllable length database 216 storing the duration of each synthesized speech syllable are stored.

  As shown in FIG. 3, the BGM database 211 includes a category column 211a in which a category indicating the type of target text to be played back is stored, a subcategory column 211b in which a subcategory of the target text is stored, and song titles in the BGM. The song title field 211c is stored, the duration field 211d is stored, and the selection number field 211e for selecting the BGM is stored. Based on the combination of the above conditions, the BGM database 211 has a plurality of BGMs having different durations suitable for the type of text to be reproduced. In a speech synthesis process to be described later, the BGM reproduced in synchronization with the target text includes the category stored in the category column 211a, the subcategory stored in the subcategory column 211b, and the duration of the BGM stored in the duration column 211d. And is determined on the condition. Further, the text for which no sub-category is specified is determined from the category and duration, and the text for which no category is specified is determined from the BGM of the all category.

  Next, the reading information database 212 will be described with reference to FIGS. The reading information is added to the target text as a reading information tag. The reading information database 212 is a database for determining the content of information using the reading information tag as a search key. FIG. 4 shows an example of a reading information database 212 of a type in which the gender, age, wording, pitch, tone, and voice quality of the reading voice are stored in a plurality of types or levels as reading information. . In this example, the reading information tag of the target text is individually specified in the form of, for example, “<sex: 2>”, “<word usage: 2>”, and the reading information is integrated with these tags. It is determined.

  FIG. 5 shows an example of a read-out information database 212 of a type composed of a combination of various read-out information. In this example, the reading information database 212 includes a selection number field 212a, a sex field 212b, an age field 212c, a wording field 212d, a pitch field 212e, a tone field 212f, a voice quality field 212g, and a default value field 212h. A combination of reading conditions stored in 212b to voice quality column 212g can be designated by a number in selection number column 212a and added to the target text as a reading information tag. When the reading information tag is not added, the one stored in the default value column 212h (G in the example of FIG. 5) can be used.

  Next, the storage area of the RAM 22 will be described with reference to FIG. FIG. 6 is a schematic diagram showing a storage area of the RAM 22. The RAM 22 is for storing the results of extracting various tags added to the target text. The RAM 22 includes an environment information storage area 221, a category information storage area 222, a subcategory information storage area 223, a format information storage area 224, a reading information storage area 225, a stop information storage area 226, a partial information storage area 227, and reading voice data. The storage area 228, the reading time storage area 229, and the selected BGM storage area 230 are arranged so that data is sequentially stored in each storage area every time the target text is processed.

  Next, the operation of the portable information terminal 1 configured as described above will be described with reference to the flowchart of FIG. FIG. 7 is a flowchart showing the overall flow of the speech synthesis process in the portable information terminal 1. Here, the text to be subjected to speech synthesis processing may be stored in advance in the non-volatile memory 21 or may be distributed from the network via infrared communication or a modem (not shown). The text data targeted in this embodiment has tag information added in advance as a condition for reproducing the BGM.

  When the process is started, first, the text to be read is analyzed to determine whether there is an environment information tag (S12). The environmental information tag is used to indicate the quietness of the environment in which sound and BGM are output. For example, the environmental information tag is added to the text in the form of “<Environment: Inside a train”, “<Environment: Late night>”, etc. ing. If there is an environmental information tag (S12: YES), the tag information is extracted (S16). When there is no environmental information tag (S12: NO), the surrounding sound is detected using the microphone 31 (S14). The loudness is detected as 80 dB, 40 dB, or the like, for example. Next, the tag information or the extracted sound volume is stored in the environment information storage area 221 of the RAM 22 (S18). The environmental information stored here is reflected in the volume of the BGM that is selected later and the volume of the reading voice.

  Next, it is determined whether or not there is a category information tag representing a large type of target text (S20). The category information tag is added to the text in the form of, for example, “<Category: Sports>”, “<Category: News>”, “<Category: Emergency>”, “<Category: Fairy tale>”. Note that the category information is not necessarily added to all texts. If there is a category information tag (S20: YES), the category information is extracted and stored in the category information storage area 222 of the RAM 22 (S22). If there is no category information tag (S20: NO), the process proceeds to S24 as it is. The category information stored here is one of the conditions for selecting the BGM together with the subcategory information described later and the calculated reading time. When the category information is not stored, the BGM used for all categories is selected.

  Next, it is determined whether or not there is a subcategory information tag indicating a small type in the category of the target text (S24). The subcategory information tag is added to the text in the form of, for example, “<subcategory: struggle>”, “<subcategory: economy>”, “<subcategory: disaster>”, “<subcategory: foreign>”. Note that the subcategory information may not be added to all texts. If there is a subcategory information tag (S24: YES), the category information is extracted and stored in the subcategory information storage area 223 of the RAM 22 (S26). If there is no subcategory information tag (S24: NO), the process proceeds directly to S28.

  Next, the format information is extracted and stored in the format information storage area 224 of the RAM 22 (S28). The format information is composed of format information tags including the number of characters of the target text, line feeds, the number of blanks, etc., and is added to the text. For example, the format information tag is added to the text in a form such as “<number of characters: 250>”, “<blank: 5>”, or the like. Moreover, when not added, the number of characters of the target text may be counted. The format information stored here is used when calculating the reading time of the target text.

  Next, it is determined whether or not there is a reading information tag that represents information related to the voice when reading the text (S30). Information related to reading speech includes volume, gender, age, pitch, tone, voice quality, and the like. The reading information tag includes, for example, “<sex: 1>”, “<age: 3>”, “<pitch: 4>”, and the like, as in the example of the reading information database 212 of FIG. As shown in the example of the reading information database 212 in FIG. 5, it is added to the text in the form of “<reading: A>” by a combination of these. In addition, when not added, a specified value is used. If there is a reading information tag (S30: YES), the reading information is extracted and stored in the reading information storage area 225 of the RAM 22 (S32). If there is no reading information tag (S32: NO), the process proceeds to S34. The reading information stored here is simultaneously output to the audio unit 34 when reading audio data to be described later is output, and is reflected when output from the speaker.

  Next, it is determined whether or not there is an ON / OFF information tag indicating whether or not the BGM is reproduced together with the target text (S34). The ON / OFF information tag is added to the text in a form such as “<BGM: ON>”, “<BGM: OFF>”, or the like. If it is not added, it is processed in the same way as the previous text. If there is an ON / OFF information tag (S34: YES), the ON / OFF information is stored in the stop information storage area 226 of the RAM 22 (S36). If there is no ON / OFF information tag (S34: NO), the process proceeds to S38. If the stored ON / OFF information is ON, the selected BGM is played back, and if it is OFF, playback is not performed.

  Next, when BGM is reproduced for a part of the target text, it is determined whether or not there is a partial information tag indicating information for specifying the ratio and the reproduction part (S38). The partial information tag is a form that surrounds the text of the part with “<BGM>” and “</ BGM>” when the BGM is to be reproduced only in the climax part of the target text, for example, and the target text When it is desired to flow BGM to the portion corresponding to 50-75% from the beginning of the text, “<BGM: 50-75%>” is added to the text at the beginning of the text. If there is (S38: YES), the partial information is stored in the partial information storage area 227 of the RAM 22 (S40) If there is no partial information tag (S38: NO), the process proceeds to S42 as it is. The BGM duration is calculated by multiplying the read partial information by the reading time calculated in the reading time calculation process described later, and the BGM that matches the calculated BGM duration is calculated as BGM. Select from the database 211. If there is no matching BGM in the BGM database 211, select a BGM with a duration longer than the reading time and fade out the BGM when reading ends. Alternatively, the reading or BGM speed may be adjusted to end simultaneously.

  Next, speech synthesis processing of the target text is performed (S42). In this speech synthesis process, an accented reading character string is generated by using a speech synthesis dictionary 215 by a well-known method, and a pause is given and stored in the reading speech data storage area 228 of the RAM 22. Next, the continuation length of the syllable represented by each read character string is calculated for the generated read character string data with accent / pause and summed up to calculate the reading time (S44). The default value of the syllable duration is stored in advance in the nonvolatile memory 21 as the syllable length database 216. The actual syllable duration is calculated by multiplying this specified value by the utterance speed count. The generation speed coefficient reflects information regarding the pitch stored in the reading information storage area 137. Next, the calculated durations of the syllables are totaled to calculate the total duration required for reading the target text, and the total duration is stored in the reading time storage area 229.

  Next, by the above processing, the category stored in the category information storage area 222, the subcategory stored in the subcategory information storage area 223, the ratio of BGM playback stored in the partial information storage area 227, the reading time storage area 229 A suitable BGM is selected from the BGM database 211 on the condition of the reading time stored in (S46). The selected BGM is stored in the selected BGM storage area 230.

  Next, it is determined whether or not the target text currently being processed is the last text to be read (S48). If it is not the last text (S48: NO), the process returns to S18, and the same environment information as the previous text is stored in the environment information storage area 221. Since the environmental information is an index indicating the quietness of the surroundings where reading and BGM reproduction will be performed from now on, common data is stored for all texts. And the process of S20-S46 is performed about the following text, and a process is repeated until it becomes the last text.

  In the case of the last text (S48: YES), it is determined whether or not the same category is continuous for a plurality of texts stored in the RAM 22 (S50). If the same category is continuous (S50: YES), a sharing process is performed (S52). Specifically, the BGM playback time obtained from the reading time of the continuous text and the partial information is totaled, and the BGM of the same category corresponding to the total time is changed, and the selected BGM storage area 230 is overwritten. Remember. If the same category is not continuous (S50: NO), the process proceeds directly to S54.

  Then, the read audio data stored in the read audio data storage area 228, the read information stored in the read information storage area 225, and the selected BGM information stored in the selected BGM storage area 230 are output to the audio unit 34, and the audio The read-out voice data is converted to analog by the unit 34 and output from the speaker 35 in synchronization with the selected BGM (S54).

  A specific example will be described below with reference to FIG. For example, assume that there are three texts to be read out. First, when the environment information tag is not added to the target text for which processing has started (S12: NO), the surrounding sound is picked up using the microphone 31 and the volume is measured (S14). Then, the measured sound volume, for example, 80 dB, is stored in the area corresponding to the first text in the environment information storage area 221 (S18). Next, when the category information tag is searched, the tag “<category: news>” is hit (S20: YES), and this information is stored in the area corresponding to the first text in the category information storage area 222 (S22). ). Next, when the subcategory information tag is searched, the tag “<subcategory: economy>” is hit (S24: YES), and this information is stored in the area corresponding to the first text in the subcategory information storage area 223 (S26). ). Then, since the format information tag was searched but did not hit, the number of characters in the text was counted and the information of 250 characters counted was stored in the area corresponding to the first text in the format information storage area 224 (S28). .

  Next, when the reading information tag is searched, the tag “<reading: G>” is hit (S30: YES), and this information is stored in the area corresponding to the first text in the reading information storage area 225. Store (S32). When the ON / OFF information tag is searched, the tag “<BGM: ON>” is hit (S34: YES), and this information is stored in the area corresponding to the first text in the stop information storage area 226. (S36). Next, when the partial information tag is searched, the tag “<BGM: 50-75%>” is hit (S38: YES), so that information is stored in the area corresponding to the first text in the partial information storage area 227. Store (S40).

  Next, speech synthesis processing is performed on the target text, and the generated character string with reading / accent / pause is stored in an area corresponding to the first text in the reading voice data storage area 228 (S42). Then, using the syllable length database 216 for the data stored in the reading voice data storage area 228, the time required for reading is calculated, and the calculated reading time of 90 seconds is calculated as the first reading time storage area 229. The area corresponding to the text is stored (S44). The information stored in the above processing, that is, category: news, subcategory: economy, playback ratio: 25% (50-75% of text), and playback duration 22.5 calculated from reading time 90 seconds. On the condition of the second, the BGM database 211 shown in FIG. 3 is searched and K of BGM is selected (S46). Here, the duration of K of BGM is 30 seconds, which is longer than the calculated playback duration, and therefore it is preferable to set the BGM K to fade out in accordance with the duration of playback when reading ends.

  Next, since this text is not the last text (S48: NO), the process returns to S18, and the same information stored in the first text in the area corresponding to the second text in the environment information storage area 221. 80 dB is stored (S18). Next, when the category information tag is searched, the tag “<category: news>” is hit (S20: YES), and this information is stored in the area corresponding to the second text in the category information storage area 222 (S22). ). Next, when the subcategory information tag is searched, the tag “<subcategory: battle>” is hit (S24: YES), and this information is stored in the area corresponding to the second text in the subcategory information storage area 223 (S26). ). Then, the number-of-characters information obtained from “<number of characters: 150>” obtained by searching the format information tag is stored in the area corresponding to the second text in the format information storage area 224 (S28).

  Next, when a reading information tag is searched, a tag “<reading: T>” is hit (S30: YES), and this information is stored in an area corresponding to the second text in the reading information storage area 225. Store (S32). Then, since the ON / OFF information tag was searched but did not hit (S34: NO), the same ON information as the first text is stored in the area corresponding to the second text in the stop information storage area 226. . Next, when the partial information tag is searched, the tag “<BGM: 20-50%>” is hit (S38: YES), so that information is stored in the area corresponding to the second text in the partial information storage area 227. Store (S40).

  Next, speech synthesis processing is performed on the target text, and the generated character string with reading / accent / pause is stored in an area corresponding to the second text in the reading voice data storage area 228 (S42). Then, the time taken to read out the data stored in the reading voice data storage area 228 is calculated using the syllable length database 216, and the calculated reading time of 110 seconds is calculated as the second time in the reading time storage area 229. The area corresponding to the text is stored (S44). Then, the information stored in the above processing, that is, category: news, subcategory: battle, playback ratio: 30% (20-50% of text) and playback duration 33 seconds calculated from reading time 110 seconds. As a condition, the BGM database 211 shown in FIG. 3 is searched and N of BGM is selected (S46).

  Next, since this text is not the last text (S48: NO), the process returns to S18, and the same information stored in the second text in the area corresponding to the third text in the environment information storage area 221. 80 dB is stored (S18). Next, when the category information tag is searched, the tag “<category: sports>” is hit (S20: YES), and this information is stored in the area corresponding to the third text in the category information storage area 222 (S22). ). Next, when the subcategory information tag is searched, the tag “<subcategory: land>” is hit (S24: YES), and this information is stored in the area corresponding to the third text in the subcategory information storage area 223 (S26). ). Then, the number-of-characters information obtained from “<number of characters: 150>” obtained by searching for the format information tag is stored in the area corresponding to the third text in the format information storage area 224 (S28).

  Next, when the reading information tag is searched, the tag “<reading: G>” is hit (S30: YES), and this information is stored in the area corresponding to the third text in the reading information storage area 225. Store (S32). Then, since the ON / OFF information tag was searched but did not hit (S34: NO), the same ON information as the second text is stored in the area corresponding to the third text in the stop information storage area 226. . Next, when the partial information tag is searched, the tag “<BGM: 40-90%>” is hit (S38: YES), so that information is stored in the area corresponding to the third text in the partial information storage area 227. Store (S40).

  Next, speech synthesis processing of the target text is performed, and the generated character string with reading / accent / pause is stored in an area corresponding to the third text in the reading voice data storage area 228 (S42). Then, using the syllable length database 216 for the data stored in the reading voice data storage area 228, the time required for reading is calculated, and the calculated reading time of 60 seconds is calculated as the third time in the reading time storage area 229. The area corresponding to the text is stored (S44). Then, the information stored by the above processing, that is, category: sports, subcategory: land, reproduction ratio: 50% (40-90% of text) and reading duration of 30 seconds calculated from reading time of 60 seconds. As a condition, the BGM database 211 shown in FIG. 3 is searched and J of BGM is selected (S46).

  Next, since this text is the last text (S48: YES), the category information storage area 222 of the RAM 22 is searched to check whether or not the same category is continuous (S50). Then, the categories of text 1 and text 2 are the same as “news” (S50: YES). Here, common processing is performed (S52). Specifically, the sum of the BGM playback times of text 1 and text 2 is 22.5 seconds + 33 seconds = 55.5 seconds. Therefore, L of BGM corresponding to the duration time is selected, and text 1 and text 2 are selected. Is continuously stored in the selected BGM storage area 230 so that the BGM L flows. Here, the duration of L of BGM is 1 minute and 40 seconds, which is longer than the calculated playback duration. Therefore, it is preferable to set the BGM L to fade out in accordance with the playback duration when reading ends. . Then, the reading voice data stored in the reading voice data storage area 228 of each text, the reading information stored in the reading information storage area 225, and the selection BGM information stored in the selection BGM storage area 230 are sequentially supplied to the audio unit 34. Then, the audio data is converted to analog by the audio unit 34 and output from the speaker 35 in synchronization with the selected BGM (S54).

  As described above, according to the mobile terminal of the present embodiment, the environment information tag, the category information tag, the subcategory information tag, the format information tag, the reading information tag, the ON / OFF information tag, the part added to the text Information tags and reading information tags are extracted and BGM suitable for the text is selected from the BGM database using these information as a composite condition, so the user can read the text comfortably while listening to the BGM appropriate for the content of the text to be read. I can listen.

  In the above embodiment, the environment information extraction process in S16 of the flowchart of FIG. 7, the category information extraction process in S22, the subcategory information extraction process in S26, the format information extraction process in S28, the reading information extraction process in S32, S36 The CPU 10 that executes the ON / OFF information extraction process in S40 and the partial information extraction process in S40 functions as the information extraction means of the present invention. Further, the CPU 10 that executes the BGM selection process in S46 of the flowchart of FIG. 7 functions as the BGM selection means of the present invention. Further, the CPU 10 that performs the output processing in synchronization with the read-out voice in S54 of the flowchart of FIG. 7 functions as the BGM reproducing means of the present invention. Further, the CPU 10 that executes the sharing process in S52 of the flowchart of FIG. 7 functions as the sharing means of the present invention.

  The speech synthesizer of the present invention can be applied to a device having a computer capable of performing speech synthesis processing and a BGM sound source.

It is a circuit block diagram of the portable information terminal 1 which shows an example of the speech synthesizer by this invention. 3 is a block diagram schematically showing a configuration of a nonvolatile memory 21. FIG. 3 is a block diagram schematically showing a configuration of a BGM database 211. FIG. 3 is a block diagram schematically showing a configuration of a reading information database 212. FIG. 3 is a block diagram schematically showing a configuration of a reading information database 212. FIG. 3 is a schematic diagram showing a storage area of a RAM 22. FIG. 4 is a flowchart showing an overall flow of speech synthesis processing in the portable information terminal 1.

Explanation of symbols

1 mobile information terminal 10 CPU
21 Non-volatile memory 34 Audio section 35 Speaker 211 BGM database

Claims (15)

In a speech synthesizer that reads out text from synthesized speech while reproducing BGM,
Storage means for storing at least BGM data classified by category and playback time;
Information extracting means for extracting additional information added to the text;
BGM selection means for selecting BGM data from the storage means based on the additional information extracted by the information extraction means;
A speech synthesizer comprising: BGM reproduction means for reproducing BGM data selected by the BGM selection means in synchronism with reading of the text.   The additional information includes category information for classifying the content of the text, reading information for designating an attribute including the size, pitch, and tone of a reading voice that reads out the text, format information indicating the number of characters and the length of the text, 2. The speech synthesizer according to claim 1, further comprising any one of environmental information indicating a loudness of a surrounding sound to be read out.   The speech synthesis apparatus according to claim 2, wherein the BGM selection unit selects BGM data based on at least one of the category information and the format information.   4. The speech synthesizer according to claim 1, wherein the BGM selection means selects one BGM data for each of the plurality of texts. A common means for extracting common features from additional information of a plurality of consecutive texts,
The speech synthesis apparatus according to claim 1, wherein the BGM selection unit selects one BGM data for a plurality of texts shared by the sharing unit.   6. The speech synthesizer according to claim 5, wherein the common unit extracts the common category information as a common feature for a plurality of texts having the common category information. The additional information includes stop information for instructing whether or not to reproduce the BGM data, or partial information indicating a ratio of a part for reproducing the BGM,
7. The speech synthesizer according to claim 1, further comprising a BGM stop control means for starting and stopping playback of the BGM based on the stop information or the partial information. In a speech synthesis method for reading out text with synthesized speech while playing back BGM,
An information extraction step of extracting additional information added to the text;
A BGM selection step of selecting BGM data for the text from a plurality of BGM data classified and stored by at least a category and a reproduction time based on the additional information extracted by the information extraction step;
A speech synthesis method comprising: a BGM reproduction step of reproducing the BGM data selected in the BGM selection step in synchronization with the reading of the text.   The additional information includes category information for classifying the content of the text, reading information for designating an attribute including the size, pitch, and tone of a reading voice that reads out the text, format information indicating the number of characters and the length of the text, 9. The speech synthesis method according to claim 8, further comprising any one of environmental information indicating a loudness of a surrounding sound to be read out.   The speech synthesis method according to claim 9, wherein, in the BGM selection step, BGM data is selected based on at least one of the category information or the format information.   11. The speech synthesis method according to claim 8, wherein in the BGM selection step, one BGM data is selected for each of the plurality of texts. It has a common process to extract common features from additional information of multiple consecutive texts,
The speech synthesis method according to claim 8 or 9, wherein, in the BGM selection step, one BGM data is selected for a plurality of texts shared in the sharing step.   13. The speech synthesis method according to claim 12, wherein, in the sharing step, the common category information is extracted as a common feature for a plurality of texts having the same category information. The additional information includes stop information for instructing whether or not to reproduce the BGM data, or partial information indicating a ratio of a part for reproducing the BGM,
The speech synthesis method according to any one of claims 8 to 13, further comprising a BGM stop control step of starting and stopping playback of the BGM based on the stop information or the partial information. A speech synthesis program for causing a computer to execute the speech synthesis method according to claim 8.

Images