JP2010091829A - Voice synthesizer, voice synthesis method and voice synthesis program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a voice synthesizer, performing voice synthesis output of text information in a short time, and configured in a comparatively inexpensive manner, and to provide a voice synthesis method and a voice synthesis program. <P>SOLUTION: This voice-synthesis method includes a step (S101) of receiving text information; a step (S106) of reading phoneme data, corresponding to the text information, with reference to a hard disk device storing phoneme data and its cache memory, and creating voice data based on the read phoneme data; a step (S108) of determining whether instruction for voice reproduction is given; and a step (S109) of voice-outputting the created voice data to a voice output means, when it is determined that a designation for voice reproduction is given, and deleting the voice data (S110) and outputting empty voice data to the voice output means, when it is determined that instruction for voice reproduction is not given. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a speech synthesizer that performs speech synthesis of text information, and more particularly to a speech synthesizer used in an in-vehicle navigation device.

  In-vehicle electronic systems, in addition to the function to output audio of content information acquired from radio broadcasts, television broadcasts, and other media, in addition to synthesizing text information acquired from outside such as radio broadcasts, and outputting this as audio Some have functions. In particular, in a navigation device, if real-time road traffic information is output as a voice, it is possible to avoid driver's side-by-side driving and improve convenience. In Europe and the like, services such as a traffic message channel (Traffic Message Channel: hereinafter referred to as TMC) that provides road traffic information as text information using subcarriers of FM radio broadcasting are in practical use. In an in-vehicle electronic system, TMC is used to synthesize and output received road traffic information.

  There have been many reports on speech synthesis. For example, Patent Document 1 relates to a unit connection type speech synthesizer, and has a sub cost including at least an access speed cost to a storage unit that stores waveform data of a speech unit, and a cost calculated including the sub cost is included. A technique for selecting a candidate satisfying a predetermined condition is disclosed. Patent Document 2 relates to a speech synthesizer, and discloses a technique for calculating distortion generated when speech units are connected to generate synthesized speech, and selecting speech units for each synthesis unit based on the distortion. Yes.

JP 2005-266010 A JP 2007-310176 A

  FIG. 1 is a diagram showing a configuration of a conventional speech synthesizer. The voice synthesizer includes a TMC receiver 20 that receives TMC included in FM broadcast, a voice synthesizer module 30, a hard disk device 40 that stores phoneme data, and a voice output unit 50. The TMC receiving unit 20 receives text information in response to an instruction from the text information extracting unit 22 that extracts text information included in the FM broadcast, the latest text information storage unit 24 that stores the latest text information, and the play button 28. And a text information transmission unit 26 that transmits to the speech synthesis module 30. The speech synthesis module 30 includes a syntax analysis unit that parses received text information, a phoneme selection unit 34 that reads phoneme data corresponding to the parsed word or phrase from the hard disk device 40, and the read phoneme. And a phoneme coupling part 36 for coupling. The voice output unit 50 converts the voice data combined with the phoneme data into an analog signal and outputs the analog signal from the speaker.

  In the speech synthesizer shown in FIG. 1, it takes time to access the storage device 40 to read phoneme data and to combine the read phoneme data. For this reason, even if a reproduction instruction is issued from the user, it takes time for the voice synthesis process, and a delay occurs in the voice output. In particular, the larger the text information, the greater the delay.

  As a method for eliminating such a delay time, it is conceivable to store all phoneme data stored in the hard disk device 40 in a work area such as a main memory in the speech synthesis module. In addition, it takes time to load phoneme data from the hard disk device 40 to the main memory, and the capacity of the main memory has to be increased, which increases the cost. Another method is to use a central processing unit (CPU) having a high processing capacity for the speech synthesis module to improve the processing speed. However, this method also uses an expensive CPU and is expensive. turn into.

  The present invention solves such a conventional problem, and can perform speech synthesis output of text information in a short time and can be configured relatively inexpensively, a speech synthesizer, a speech synthesis method, and An object is to provide a speech synthesis program.

A speech synthesizer according to the present invention includes an acquisition unit that acquires text information from outside,
First storage means for storing phoneme data necessary for speech synthesis; phoneme data that is the same as the phoneme data read in the first storage means can be stored; and more accessible than the first storage means The phoneme data corresponding to the text information is read out from at least one of the first and second storage means with reference to the second storage means that is fast in time and the first and second storage means. It determines whether or not there is a voice reproduction instruction from the input means for creating voice data based on phoneme data, a voice output means for outputting voice, an input means for inputting voice reproduction instructions for text information, and the input means. When the voice reproduction instruction is given, the voice data created by the creation means is output to the voice output means. When there is no voice reproduction instruction, the voice data is created by the creation means. Remove the voice data, and a control means for outputting an empty audio data to the audio output means.

  Preferably, the control means instructs the creation means to create voice data corresponding to the acquired text information in response to the acquisition of text information by the acquisition means. Preferably, the control means causes the acquisition means to transmit text information to the creation means when a predetermined time has elapsed since the creation of the voice data by the creation means, and the voice corresponding to the text information transmitted to the creation means. Command the creation of data. Preferably, the control means causes the obtaining means to transmit text information to the creation means in response to a voice reproduction instruction from the input means, and instructs the creation means to create the voice data transmitted.

  Preferably, the first storage means is a mass storage device, the second storage means is a cache memory of the first storage means, and the creation means caches phoneme data corresponding to the text information. The phoneme data that was read from the memory and did not hit in the cache memory is read from the mass storage device. Preferably, the acquisition means receives a broadcast wave and extracts text information included in the broadcast wave.

  A speech synthesis method or program according to the present invention includes a step of acquiring text information, a memory for storing phoneme data necessary for speech synthesis, and a cache memory capable of storing the same phoneme data as the phoneme data read from the memory To read phoneme data corresponding to the acquired text information from at least one of the memory and the cache memory, to create audio data based on the read phoneme data, and to determine whether there is an instruction to play audio And when the voice output instruction is determined to be output, the generated voice data is output to a voice output unit, and when it is determined that there is no voice playback instruction, the generated voice data is deleted. A step of causing the sound output means to output sound of empty sound data.

  According to the present invention, the same phoneme data as the phoneme data read in the first storage device can be stored, and the voice data is created with reference to the second storage device having a fast access time. The time from the voice reproduction instruction to the voice output can be shortened. Also, when there is no voice playback instruction, the voice data is not actually output as a voice, so that the voice playback instruction that occurs irregularly can be dealt with by creating the voice data at a certain frequency without causing discomfort to the user. Can do.

  The best mode for carrying out the present invention will be described in detail with reference to the drawings.

FIG. 2 is a block diagram showing the configuration of the speech synthesizer according to the embodiment of the present invention.
The speech synthesizer 100 according to the present embodiment includes an input unit 110 that inputs an instruction from a user, a reception unit 120 that receives text information from the outside, and speech that creates speech data for speech synthesis output of the text information. A synthesis module 130, a memory 140 for storing voice data and the like for voice synthesis output, a voice output unit 150 for outputting voice based on the voice data, and an external I / F 160 forming an interface for connecting to an external device And an internal bus 170 for connecting them.

  The input unit 110 includes input devices such as a remote controller and a mouse. The user can instruct voice reproduction of the text information via the input unit 110. The receiving unit 120 receives text information from radio broadcast, television broadcast, and other media. For example, the road traffic information superimposed on the subcarrier of FM broadcasting like TMC is received. When receiving road traffic information such as TMC, the receiving unit 120 always receives FM broadcasts and extracts the latest road traffic information.

  The speech synthesis module 130 creates speech data corresponding to the text information received by the receiving unit 120 and stores the created speech data in the memory 140. When a voice reproduction instruction is given from the voice synthesis module 130, the voice output unit 150 converts voice data stored in the memory 140 into an analog signal and outputs the analog signal from the speaker.

  FIG. 3 shows the configuration of the receiving unit 120. The reception unit 120 includes a text information extraction unit 121, a text information transfer unit 122, and a text information storage unit 123. The text information extraction unit 121 extracts text information from the FM broadcast as described above. The text information transfer unit 122 is preferably configured so that the receiving unit 120 receives the text information (Condition 1), receives a voice playback instruction from the user (Condition 2), or creates a certain amount of time after generating the voice data. When the time has elapsed (condition 3), the text information is transferred to the speech synthesis module 130. Whether the condition 2 or 3 is satisfied is determined by the speech synthesis module 130, and the speech synthesis module 130 makes a transmission request to the receiving unit 120. The text information storage unit 123 stores the latest received text information and the previous text information, and deletes older text information.

  FIG. 4 shows the internal configuration of the speech synthesis module. The speech synthesis module 130 includes a speech data creation device 200 that creates speech data, a phoneme data storage device 210 that stores phoneme data, a speech synthesis device 200 that controls the speech data creation device 200, the reception unit 120, the speech output unit 150, and the like. And a control device 220.

  In response to a command from the speech synthesis control device 220, the speech creation data device 200 reads the necessary phoneme data corresponding to the text information from the phoneme data storage device 210, and creates the speech data by combining the read phoneme data. To do. The phoneme data storage device 210 includes a large-capacity hard disk device 212 and a cache memory 214 that store phoneme data. When searching for phoneme data, the voice data creation device 200 accesses the cache memory 214, and if there is phoneme data hitting the cache memory 214, reads the phoneme data and if there is no phoneme data hitting the cache memory 214. The phoneme data is retrieved from the hard disk device 212, and the phoneme data is read out.

  The cache memory 214 stores the same phoneme data as the phoneme data read by the hard disk device 212, thereby reducing the time required for searching for phoneme data. However, since the storage capacity of the cache memory 214 is limited, the oldest phoneme data may be overwritten in order, or the earliest accessed phoneme data may be rewritten according to an algorithm such as LRU (Least Recently Used). . The cache memory 214 is composed of a memory, such as an SRAM, whose access time is faster than that of the hard disk device 212.

  FIG. 5 shows the configuration of the audio data creation device 200. The speech data creation device 200 accesses the phoneme data storage device 210 based on the result of the syntax analysis, the text sequence acquisition unit 201 that acquires the text sequence from the reception unit 120, the syntax analysis unit 202 that performs syntax analysis of the acquired text sequence. Then, a phoneme selection unit 203 that selects phoneme data therefrom, a phoneme combination unit 204 that combines the selected phoneme data to create speech data, and the created phoneme data is transmitted to a specified area of the memory 140. An audio data transmission unit 205 is included.

  The syntax analysis unit 202 analyzes a series of text strings into a subject, a predicate, a particle or the like as a word or a phrase. The phoneme selection unit 203 reads phoneme data corresponding to the phonemes included in the analyzed word or phrase from the phoneme data storage device 210, and the phoneme combination unit 204 combines these phoneme data. The voice data is a combination of phoneme data, and the created voice data is written into the memory 140 by the voice data transmission unit 205 in response to a command from the voice synthesis control device 220.

  FIG. 6 shows a functional block diagram of the speech synthesis control device. The voice synthesis control device 220 includes a text information transfer request unit 221 that requests text information transfer, a voice instruction determination unit 222 that determines whether or not a voice reproduction instruction has been received from the input unit 110, and a voice data creation device 200. The voice data creation request unit 223 that requests creation of voice data, the voice data transmission request unit 224 that requests the voice creation data device 200 to send voice data, and the pseudo-sound playback are stored in the memory 140. The audio data deleting unit 225 for deleting the audio data and the audio output requesting unit 226 for requesting the audio output unit 150 to output the audio data. The speech synthesis control device 220 may include, for example, a microcontroller, a microcomputer, or a microprocessor, and these may execute a program to perform the above functions.

  Next, the operation of the speech synthesizer according to the present embodiment will be described with reference to the flowchart shown in FIG. As described above, when receiving the latest text information, the receiving unit 120 transmits the text information to the speech synthesis module 130. When receiving the text information (step S101), the speech synthesis control device 220 instructs the speech data creation device to create speech data (step S102). In addition to this, when the speech synthesis control device 220 determines that the playback instruction determination unit 222 has instructed the playback of the voice (condition 2), or the voice data creation request unit 223 transmits a creation command. When a certain period of time has passed (condition 3), the text information transfer request unit 221 requests the receiving unit 120 to transmit text information. In response to this transmission request, the receiving unit 120 transmits text information. When the text information is received by the speech synthesis module 130 (step S101), a voice data creation command is issued (step S102).

  Next, when the text string acquisition unit 201 of the voice data creation device 200 acquires text information (step S103), the syntax analysis unit 202 performs syntax analysis of the text information (step S104). Next, the phoneme selection unit 203 accesses the storage device 210 and reads out phoneme data corresponding to the parsed word or phrase from the storage device 210. Then, the phoneme combination unit 204 combines the read phoneme data (step S105) to create voice data (step S106). When reading the phoneme data, the same phoneme data as the phoneme data read from the hard disk device 212 is stored in the cache memory 214.

  Next, the voice data transmission request unit 224 of the voice synthesis control device 220 requests the voice data creation device 200 to send voice data. This request includes address information of the memory 140 where the audio data is to be stored. The voice data transmission unit 205 writes the created voice data in the memory 140 in response to the voice data transmission request (step S107).

  Next, the voice instruction determination unit 222 of the voice synthesis control device 220 determines whether or not there has been a voice reproduction instruction from the user (step S108), and when there is a reproduction instruction, the voice output request unit 226 The audio output unit 150 is requested to output audio and the audio data stored in the memory 140 is transmitted to the audio output unit 150 (step S109). The audio output unit 150 converts the transmitted audio data into an analog signal and outputs the audio from the speaker.

  On the other hand, when there is no voice reproduction instruction from the user, the voice data deleting unit 225 deletes the voice data stored in the memory 140 (step S110), and the voice output requesting unit 226 outputs the voice in the same manner as described above. An audio output request is made to the unit 150. In this case, since the audio data is deleted from the memory 140, the audio output unit 150 performs the audio output of empty data, so that the audio output is practically not performed, so-called pseudo reproduction. The voice synthesis control device 220 determines whether or not the voice reproduction processing has ended, and if not, continues the processing from step S105 (step S111). By including such pseudo reproduction processing, voice data corresponding to the text information is created even if there is no voice reproduction instruction from the user, so that the phoneme data having a high hit rate is stored in the cache memory 214. Can be kept. Then, if there is an audio reproduction instruction from the user, the audio data creation device 200 creates audio data. If the text information to be reproduced is the same as the pseudo-reproduced text information, the cache data is referred to. The time for reading phoneme data can be greatly shortened.

  FIG. 8 is a comparative example of processing time between the speech synthesizer according to the present embodiment and the conventional speech synthesizer. When “TEST” and “FAST” were output as input texts, the time until the speech output by the conventional speech synthesizer could be reduced by 91.20% and 85.70%.

  FIG. 9 is a configuration diagram when the speech synthesizer according to the present embodiment is applied to a navigation device. The vehicle-mounted navigation device 300 includes a tuner 310 that receives radio broadcasts and television broadcasts. By extracting text information (road traffic information) superimposed on the radio broadcasts and television broadcasts, the text information is synthesized with speech. The apparatus 100 can output a sound. Further, when the navigation device 300 includes the wireless communication unit 320, the text information received by the wireless communication unit can be output by the speech synthesizer 100.

  In the above embodiment, the example in which the phoneme data storage device 210 is included in the speech synthesis module 130 has been shown. However, the present invention is not limited thereto, and the storage device 210A may be connected to the internal bus 170 as shown in FIG. . In this case, the speech synthesis module reads phoneme data via the internal bus 170. The hard disk device included in the storage device 210A may store not only phoneme data but also other data. For example, when combined with a navigation device as shown in FIG. 9, the hard disk device may store map data or the like. Further, if the access time to the memory 140 via the internal bus 170 is faster than the storage device 210A (hard disk device), the memory 140 can be used as a cache memory.

  Furthermore, in the above-described embodiment, the voice synthesis control device 220 deletes the voice data. However, the present invention is not limited to this, and the voice data creation device 200 determines that the voice synthesis control device 220 has not issued a voice playback instruction. Audio data may be deleted. In this case, the voice data transmission unit 205 of the voice data creation apparatus 200 transmits empty voice data to the voice synthesis control apparatus 220.

  Further, in the above embodiment, when performing audio output, the audio data stored in the area of the memory 140 is transmitted to the audio output unit 150, but the address where the audio data is stored is constant. If so, it is not always necessary to transmit audio data. In this case, the audio output unit 150 can read out audio data from a predetermined address in the memory 140 and reproduce it.

  Further, in the above-described embodiment, the example in which the audio data is stored in the memory 140 has been described, but the audio data is not necessarily stored in the memory 140. For example, the audio output unit 150 may include a ring buffer for audio output and store audio data in the ring buffer. When performing pseudo reproduction, the audio data stored in the ring buffer is emptied. When the ring buffer is used, the audio data is stored so that the old audio data is sequentially overwritten, while the audio output unit 150 reproduces the audio in order from the old audio data.

  Furthermore, in the said Example, although the road traffic information received irregularly as text information was illustrated, text information may be other information not only road traffic information. For example, the text information may be an email or other text document.

  The preferred embodiment of the present invention has been described in detail above, but the present invention is not limited to the specific embodiment, and various modifications can be made within the scope of the present invention described in the claims. Deformation / change is possible.

It is a block diagram which shows the structure of the conventional speech synthesizer. It is a block diagram which shows the structure of the speech synthesizer based on the Example of this invention. It is a block diagram which shows the internal structure of the receiving part shown in FIG. It is a block diagram which shows the structure of the speech synthesis module shown in FIG. It is a block diagram which shows the structure of the audio | voice data production apparatus shown in FIG. FIG. 5 is a functional block diagram of the speech synthesis control device shown in FIG. 4. It is an operation | movement flowchart of the speech synthesizer concerning a present Example. It is a figure which shows the example which contrasted the time required for the audio | voice reproduction | regeneration by the speech synthesizer concerning a present Example with the conventional apparatus. It is the figure which applied the speech synthesizer of a present Example to the navigation apparatus. It is a figure which shows the example of the other speech synthesizer of a present Example.

Explanation of symbols

100: Speech synthesis device 110: Input unit 120: Reception unit 130: Speech synthesis module 140: Memory 150: Speech output unit 160: External IF 170: Internal bus

Claims (9)

An acquisition means for acquiring text information from the outside;
First storage means for storing phoneme data necessary for speech synthesis;
Second storage means capable of storing the same phoneme data as the phoneme data read in the first storage means and having a faster access time than the first storage means;
Creation of referring to the first and second storage means, reading out phoneme data corresponding to the text information from at least one of the first and second storage means, and generating voice data based on the read phoneme data Means,
Audio output means for outputting audio;
An input means for inputting an instruction to reproduce audio of the text information;
The presence or absence of an audio playback instruction from the input means is determined, and when there is an audio playback instruction, the audio data created by the creating means is output to the audio output means,
Control means for deleting the voice data created by the creation means and outputting empty voice data to the voice output means when there is no voice reproduction instruction;
A speech synthesizer. The speech synthesizer according to claim 1, wherein the control means instructs the creation means to create speech data corresponding to the acquired text information in response to the acquisition of text information by the acquisition means. The control means causes the acquisition means to transmit text information to the creation means when a predetermined time has elapsed since the creation of the voice data by the creation means, and the voice data corresponding to the text information transmitted to the creation means. The speech synthesizer according to claim 1 which instructs creation. The control means causes the acquisition means to transmit text information to the creation means in response to a voice reproduction instruction from the input means, and instructs the creation means to create voice data transmitted thereto. The speech synthesizer according to 1. The first storage means is a mass storage device, the second storage means is a cache memory of the first storage means, and the creation means sends phoneme data corresponding to the text information from the cache memory. The speech synthesizer according to claim 1, wherein the phoneme data that has not been read and hit in the cache memory is read from the mass storage device. The speech synthesis apparatus according to claim 1, wherein the acquisition unit receives a broadcast wave and extracts text information included in the broadcast wave. The speech synthesizer according to claim 6, wherein the speech synthesizer is mounted on a vehicle, and the text information is road traffic information. A speech synthesis method for synthesizing speech based on speech data,
Obtaining text information;
Reference is made to a memory for storing phoneme data necessary for speech synthesis and a cache memory capable of storing the same phoneme data as the phoneme data read from the memory, and the phoneme data corresponding to the acquired text information is stored in the memory and the cache memory. Reading from at least one;
Creating audio data based on the read phoneme data;
Determining whether or not there is an instruction to play audio;
When it is determined that there is an instruction for audio reproduction, the generated audio data is output to the audio output means, and when it is determined that there is no instruction for audio reproduction, the generated audio data is deleted and empty audio data Voice output to the voice output means;
A speech synthesis method comprising: A speech synthesis program for synthesizing speech based on speech data,
Obtaining text information;
Reference is made to a memory for storing phoneme data necessary for speech synthesis and a cache memory capable of storing the same phoneme data as the phoneme data read from the memory, and the phoneme data corresponding to the acquired text information is stored in the memory and the cache memory. Reading from at least one;
Creating audio data based on the read phoneme data;
Determining whether or not there is an instruction to play audio;
When it is determined that there is an instruction for audio reproduction, the generated audio data is output to the audio output means, and when it is determined that there is no instruction for audio reproduction, the generated audio data is deleted and empty audio data Voice output to the voice output means;
A speech synthesis program.

Images