JP4653572B2 - Client terminal, speech synthesis information processing server, client terminal program, speech synthesis information processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To generate a good-quality synthesized voice close to a real voice on a mobile terminal with limited processing capacity. <P>SOLUTION: The text analysis, the prosodic parameter calculation, and the voice fragment search processing which require major processing capacity are conducted on a server side. A client terminal reads out voice fragment data from a voice fragment database provided to the client terminal based on the voice fragment information sent from the server, connects the voice fragment data read out sequentially and generates synthesized voice data, and outputs the synthesized voice data as synthesized voices sequentially. <P>COPYRIGHT: (C)2007,JPO&amp;INPIT

Description

  The present invention relates to a speech synthesis method, a speech synthesis information processing method, a speech synthesis method, a client terminal that operates using the speech synthesis information processing method, and a speech synthesis information processing server.

In recent years, along with a decrease in the cost of using a large-capacity storage device, tens of minutes or more of large-capacity speech data is stored in a large-capacity recording device as it is, and speech segments are generated according to input text and prosodic information. A waveform-connected corpus-based speech synthesis method that synthesizes high-quality speech by appropriately selecting, connecting and transforming is proposed (Patent Document 1, Non-Patent Document 1).
By such a method, it has become possible to generate a high-quality synthesized speech that is physically equivalent to the real voice. Specifically, a speech unit that is partially or completely matched with a phoneme sequence corresponding to a character string to be synthesized is searched from a speech database using a speech unit dictionary composed of a binary tree or the like. The appropriate combination of speech units is selected by a method such as DP (Dynamic Programming) from a number of speech units that are costed according to a combination-based evaluation scale for evaluating the similarity of The speech synthesis is performed by connecting the selected speech segments in order (Non-Patent Document 2). However, in such a speech synthesis method, a high-quality synthesized speech cannot be generated unless an appropriate speech segment exists in the speech database.

Therefore, in order to synthesize various texts with high quality, it is indispensable to use a speech database containing a variety of speech segment variations. Therefore, in recent years, the number of speech segment variations has been increased and high quality speech synthesis has been achieved. Therefore, development is progressing in a direction to further increase the capacity of the voice database. However, although the quality of the synthesized speech is improved by increasing the capacity of the speech database, the number of speech units stored in the speech database naturally increases, so that depending on the input text during speech synthesis The amount of search processing required to search for an appropriate speech segment from a huge number included in the speech database is increasing.
Japanese Patent No. 2761552 M. Beutnagel, A. Conkie, J. Schoroeter, Y. Stylianou, and A. Sydral, "Choose the best to modify the least: A new generation concatenative synthesis system", in Proc. Eurospeech'99, 1999, pp. 2291 -2294 “Waveform Selection Method in Waveform Editing Type Rule Synthesis Method”, Hirokawa et al., IEICE Technical Report, SP89-114, pp.362-369 (1990)

The waveform-connected corpus-based speech synthesis method can generate high-quality synthesized speech by increasing the capacity of the speech database, but on the other hand, the amount of processing increases compared to the conventional method. It is coming. However, personal computers, workstations, and the like that have mainly required speech synthesis functions so far have developed remarkable processing capabilities and storage capacities in recent years.
On the other hand, in the field of small devices such as mobile phones, car navigation systems, and home appliances, various software that is easy to use for users has been developed in order to lead to further spread and differentiation of devices. It is beginning to be considered a necessary technology because it allows easy-to-understand voice information transmission.

However, in such devices, the processing capacity and the amount of memory are often very low, and even if that is not the case, most of the processing capability and memory are such as image display and device control that are the main applications of those devices There is almost no memory and processing power available for speech synthesis.
In order to reduce the amount of calculation processing, it is possible to speed up to some extent by simplifying the calculation processing or performing branch hunting processing at the time of speech unit search, but in that case the optimum unit is not selected and synthesized There is a possibility that the quality of the voice is deteriorated. There is also a method to reduce the search processing time by securing the speech unit index etc. in the memory, but in that case a large amount of external memory for work is required, but such external memory can be installed Often not. Therefore, until now, it has been very difficult to operate the waveform-connected corpus-based speech synthesis method with these devices.

  For this reason, as a method of outputting high-quality synthesized speech from a mobile device, a client / server configuration is used, text is transmitted from a mobile device that is a client, and synthesized speech from text is transmitted from a high-performance workstation that is a server. Generate and transmit synthesized speech to the client, or text analysis processing that requires a small amount of processing and a small amount of memory on the client mobile device, and then sends the text analysis result to the server to generate a large amount of speech A method to solve the above problem easily is also conceived by operating only the speech synthesizer that requires segment data and work memory on the server, generating synthesized speech from the text analysis result, and sending the synthesized speech to the client. It is done. However, in these cases, the synthesized voice corresponding to the text data to be synthesized is generated and transmitted on the server side, and no voice can be output from the client until it is received by the client side, so depending on the speed and throughput of the network Has a drawback that it takes a very long response time until sound is output.

  In addition, at present, mobile phone packet networks are often used for mobile phones and car navigation systems, but in that case, the usage fee for the network becomes very high because charging is done on a pay-as-you-go basis. There is also. Furthermore, the packet network has problems such as the voice being interrupted in the middle because the delay and speed fluctuation of the network itself are extremely large.

Analyze the text you want to synthesize, determine the appropriate speech segment sequence using the speech segment index based on the text analysis results, and then send at least speech segment sequence information from the information required for speech synthesis A speech synthesis information processing server having a function is prepared.
The client side that wants to output synthesized speech is provided with at least a function for receiving speech unit sequence information, a speech unit database storing speech units, and a speech synthesis unit.
In addition, both the speech synthesis information processing server and the client have a data transmission / reception function using a network, and the client and the speech synthesis information processing server are connected via the network.

  A server that can be used as a high-performance workstation such as a text server that requires a large amount of processing and a large amount of memory to determine a speech segment sequence, with a client-server configuration connected via a network in this way. To do. In addition, the speech unit database is stored in a memory dedicated to the read memory or a storage device, and the processing and reading from the speech unit database that requires only a small amount of processing and a small amount of memory are performed on the client side with less work memory. To do.

  A speech synthesis information processing server having a function of determining a speech unit sequence necessary for speech synthesis from text and transmitting information of the determined speech unit sequence on a network, and a speech unit storing the speech unit When preparing a client having a speech synthesis unit with a database and using speech synthesis in the client, information such as speech segment sequence information and prosodic parameters corresponding to the text to be synthesized is received from the speech synthesis information processing server, Speech unit data corresponding to the received speech unit sequence information is read from the speech unit database, and speech synthesis processing is performed using the speech unit data to generate high-quality synthesized speech. By performing processing in this way, high-quality and high-speed processing is possible regardless of the processing performance and the type of network under the condition that the processing performance on the client side is limited in advance and a network using limited speed or usage is used. Realization of synthesized speech.

This is because the speech segment sequence received from the speech synthesis information processing server is sufficient if the client side has at least processing performance and memory to the extent that speech synthesis is possible using the speech segment read and the speech segment read. Can be synthesized by reading and synthesizing a speech unit from a speech unit database.
Also, since the speech synthesis process is executed on the client side, speech synthesis can be output regardless of the crowded listing on the network.
Furthermore, since the speech unit sequence information and prosodic parameters to be transmitted using the network have a very small amount of data compared to the speech data, the charge can be reduced even if the mobile packet network is used. . Also, when sending synthesized speech, it takes a long time to send a large number of packets, and there is a high possibility that a delay will occur in the meantime. If a delay occurs, the speech will be interrupted. Therefore, it is possible to extremely reduce the occurrence of a situation where the synthesized speech is interrupted midway.

The client terminal that operates according to the speech synthesis method according to the present invention and the speech synthesis information processing server that operates according to the speech synthesis information processing method can both be configured by hardware, but the present invention proposes the simplest implementation. An embodiment in which a client terminal program and a speech synthesis information processing program are installed in a computer and the computer functions as a client terminal or a speech synthesis information processing server is the best.
When the computer functions as a client terminal, the computer constructs at least a text transmission unit, a speech unit information receiving unit, a speech unit data reading unit, a speech unit connection unit, and a speech output unit by a program. And function as a client terminal for speech synthesis.

  When the computer is caused to function as a speech synthesis information processing server, the computer has at least a text data reception unit, a text analysis unit, a prosodic parameter acquisition unit, a speech unit index, a speech unit search unit, and a speech unit, according to a program. A single information transmission unit is constructed and functions as a speech synthesis information processing server.

A first embodiment of the present invention will be described below.
First, FIG. 1 shows a conceptual diagram of the entire system. The client terminal 1 uses synthesized speech. When speech synthesis is executed by a speech synthesizer or a speech synthesis program in the client terminal 1, the speech synthesis information processing server 3 is connected via the network 2. Examples of the network 2 include a cellular phone packet communication network, ADSL using a telephone line, and FTTH using an optical fiber. Although only one client terminal is shown in FIG. 1, a situation where a plurality of client terminals 1 simultaneously access the speech synthesis information processing server 3 can be considered.

The speech synthesis information processing server 3 determines speech unit information corresponding to the text to be synthesized and transmits it to the client terminal 1. The client terminal 1 receives the speech unit sequence information necessary for speech synthesis from the speech synthesis information processing server 3, and the speech synthesizer or speech synthesis program in the client terminal 1 uses the received speech unit information to perform speech. Perform the synthesis.
Here, the speech synthesizer in the client terminal 1 may be configured by causing a known computer including a CPU (Central Processing Unit), a RAM, a hard disk device, and the like to execute a predetermined program. 2, a work memory 40 composed of a RAM or the like for storing a program and computation results, an MPU (Micro Processing Unit) 41 for performing computation based on the program and controlling each component of the speech synthesizer, speech A storage memory 42 composed of a ROM or the like for storing fragment data and other files, a data transmission / reception unit 43 for transmitting text data to the network 2 and receiving data from the network 2, and an audio output unit 44 are provided. To do.

In a personal computer or the like, the storage memory 42 may be implemented with a magnetic disk or the like. The speech synthesis information processing server 3 is configured by causing a known computer including, for example, a CPU, a RAM, and a hard disk device to execute a predetermined program.
FIG. 3 is an illustration of a conceptual configuration diagram of the speech synthesis information processing server 3 in the present embodiment.
The speech synthesis information processing server 3 of the present embodiment includes a text data receiving unit 9, a text analyzing unit 10, a prosody parameter acquiring unit 11, a speech unit searching unit 12, a speech unit information sending unit 13, and a speech unit index Index- 1 and.

FIG. 4 is a flowchart for explaining speech unit information transmission processing in the speech synthesis information processing server 3 in this embodiment.
Details of speech unit information transmission according to the present embodiment will be described below with reference to FIGS. 3 and 4 described above. When the speech synthesis information processing server 3 receives the text data sent from the client terminal 1 (step S4-1), the text analysis unit 10 performs text analysis processing to generate reading information and prosodic information ( S4-2). The text analysis processing here is mainly composed of morphological analysis processing and reading / accenting processing. However, there are various methods for these processing methods. For example, (Reference: Japanese Patent No. 3379634, “Morphology”). Analysis method and recording medium on which morphological analysis program is recorded ") and (reference document: Japanese Patent No. 3518340" Reading Prosody Information Setting Method and Apparatus and Recording Medium on which Reading Prosody Information Setting Program is Stored ") Can also be done based on.

  Next, the prosodic parameter acquisition unit 11 obtains prosodic parameters based on the prosodic information (S4-3). Here, the prosodic parameters include pitch (fundamental frequency), phoneme duration length, and the like, but there are also methods for obtaining them, for example (Reference: Japanese Patent No. 3240691 “Pitch pattern generation method, apparatus thereof” In addition, the pitch (fundamental frequency) can be obtained by the method of (Reference: Japanese Patent No. 3344487, “Audio basic frequency pattern generation device”). Also, for example (reference: Ukiki et al., “Control of vowel duration using linguistic information” vol.75, No.3 pp.467-463, Theory of Science, 1992) and (Reference: MD Riley. “Tree-based modeling for speech synthesis.” In G. Bailly C. Benoit, and TR Sawallis, editors, Talking Machines: Theories, Models, and Designs, pages 265-273. Elsevier, 1992.) The duration time can also be obtained.

Next, according to the above-described reading information and prosodic parameters, the speech unit search unit 12 determines an optimal speech unit sequence using the speech unit index Index-1 (S4-4). Here, the speech unit index configuration and speech unit sequence determination method include, for example, Japanese Patent No. 3515406 specification “speech synthesis method and apparatus”. Here, the speech unit index shown in FIG. This will be described using the conceptual diagram.
Using the set of reading information and prosodic information as a key, a speech unit index Index-1 is searched, and from the corresponding speech unit index Index-1, a phoneme sequence and prosody that are reading information belonging to a similar range of the set are searched. A speech unit corresponding to the set of parameters is selected to determine a speech unit sequence.

  Here, the similarity range is a concept including, for example, a case where reading information and prosodic parameters completely match, a portion of which matches, and a high similarity specified by cost. For example, when the phoneme is “a” as the reading information, the preceding phoneme environment is “#”, and the prosodic parameters are given the condition that the average F0 is 200 ± 10 Hz, the speech unit as shown in FIG. Three of A1, A2 and A3 will fit. Further, with respect to the reading information and the prosodic parameter, a total cost value is calculated from the reading information and the prosodic parameter suitable for the speech element in the speech element index Index-1, and the speech element that minimizes the cost is selected. You can also.

Here, although it is a total cost calculation method, for example, the total cost Pnew can be obtained by using a sub-cost function as described below (reference: “a waveform selection method considering spectrum continuity in a waveform editing type composition method) "The Acoustical Society of Japan, 2-6-10, pp.413-414, 1990/9)".
The number of phonemes in which the phoneme sequence as the reading information matches the phoneme sequence as the reading information of the speech unit is n, and the sub-cost function corresponding to the reading information is
C ₁ (n) = 1 / e ⁿ
Among the prosodic parameters, sub-cost functions corresponding to the average pitch Vp and the average pitch Vs of speech segments
C ₂ (Vp, Vs) = | Vp-Vs | ²
Of the prosodic parameters, sub-cost functions corresponding to the pitch slope Fp and the pitch slope Fs of the speech segment
C ₃ (Fp, Fs) = | Fp-Fs | ²
Among the prosodic parameters, sub-cost functions corresponding to the time length Tp and the time length Ts of the speech unit
C ₄ (Tp, Ts) = | Tp-Ts | ²
The sub-cost function corresponding to the amplitude Ap of the prosodic parameters and the amplitude As of the speech segment
C ₅ (Ap, As) = | Ap-As | ²
When ω ₁ , ω ₂ , ω ₃ , ω ₄ , and ω ₅ are given in advance as sub cost weights corresponding to the sub cost functions of C ₁ , C ₂ , C ₃ , C ₄ , and C ₅ , respectively. Ω = ω ₂ C ₂ (Vp, Vs) + ω ₃ C ₃ (Fp, Fs) + ω ₄ C ₄ (Tp, Ts) + ω ₅ C ₅ (Ap, As)
P = ω ₁ C ₁ (n) + (1-ω ₁ ) Ω
Pnew = (1 + G) P: G is an acoustic scale. By using the total cost of each speech segment thus obtained, for example, using a general DP (Dynamic Programming) method or Viterbi method, it is easy. Speech segments that minimize the cost can be sequentially selected, and speech segment sequences can be determined.

Next, the speech unit information sending unit 13 transmits speech unit information to the client terminal 1 in the speech unit series (S4-5).
FIG. 5 is an illustration of a conceptual configuration diagram of the client terminal 1 in the embodiment. The client terminal 1 of the present embodiment includes a text data transmitting unit 29, a speech unit information receiving unit 30, a speech unit data reading unit 31, a speech unit connection unit 32, a speech output unit 33, and a speech unit database DB-1. And have.
FIG. 6 is a flowchart for explaining speech synthesis processing in the client terminal 1 in the present embodiment.

The details of the speech synthesis process in the present embodiment will be described below with reference to FIGS.
First, the client terminal 1 transmits text data corresponding to synthesized speech to be generated to the speech synthesis information processing server 3 through the network 2 (S6-1).
The speech synthesis information processing server 3 analyzes the sent text data and returns speech unit information to the client terminal 1. The client terminal 1 receives the speech unit information transmitted from the speech synthesis information processing server 3 through the network 2 by the speech unit information receiving unit 30 (S6-2).

Next, the speech unit data reading unit 31 reads speech unit data from the speech unit database DB-1 based on the received speech unit information (S6-3).
Here, the speech unit index Index-1 existing in the speech synthesis information processing server 3 storing the received speech unit information, and the speech unit database provided in the client terminal 1 as shown in FIG. Even if DB-1 physically exists separately, DB-1 can be read out easily from the speech unit information because it is theoretically associated.
For example, as speech unit information, from speech unit A2, speech unit R1, speech unit I2,..., File as speech unit data based on speech unit storage information corresponding to speech unit A2. Voice data of number 8, start point 10 msec, time length 110 msec is read out, and voice unit data corresponding to file number 23, start point 5225 msec, time length 15 msec, voice unit I2 as voice unit data corresponding to voice unit R1 Then, the speech unit data is sequentially read out such that the audio data having the file number 23, the start point 5240 msec, and the time length 95 msec is read out.

Next, in the speech unit connection unit 32, the read speech units are sequentially connected to generate synthesized speech data (S6-4). Here, the speech unit data may be simply connected in the order of time, but it is also easy to supplement between different speech units in terms of time or frequency. (Reference document: Japanese Laid-Open Patent Publication No. 07-072897 "Speech Synthesis Method and Device") Finally, the connected speech segment data is output as synthesized speech in the speech output unit 33 (S6-5).
In the above description, the text data is described as being transmitted from the client terminal 1 to the speech synthesis information processing server 3. However, this is not always necessary. For example, a server having a large number of text data on a network is prepared. The target text data may be sent to the speech synthesis information processing server 3 by instructing the client terminal 1 to send desired text data to the speech synthesis information processing server 3.

FIG. 7 is an illustration of a conceptual configuration diagram of the speech synthesis information processing server 3 ′ proposed in the second embodiment of the present invention.
The speech synthesis information processing server 3 ′ of the present embodiment includes a text data receiving unit 9, a text analysis unit 10, a prosody parameter acquisition unit 11, a speech unit search unit 12, a speech unit information / prosodic parameter transmission unit 14, and a speech unit. And an index Index-1.
FIG. 8 is a flowchart for explaining speech element information transmission processing in the speech synthesis information processing server 3 ′ proposed in the second embodiment.

The details of the speech unit information transmission according to the second embodiment will be described below with reference to FIGS. The speech synthesis information processing server 3 ′ receives the text information transmitted from the client terminal until the speech segment information is obtained until the text analysis unit 10, the prosodic parameter acquisition unit 11, and the speech segment search unit 12. The contents of the configuration and processing can be executed in the same manner as in the first embodiment.
In the second embodiment, the following processing is performed in order to improve the synthesized speech quality at the client terminal.

The speech unit information / prosodic parameter transmission unit 14 transmits the prosody parameters obtained by the prosody parameter acquisition unit 11 to the client terminal in addition to the speech unit information determined by the speech unit search unit 12. .
FIG. 9 is an illustration of a conceptual configuration diagram of the client terminal 1 ′ corresponding to the speech synthesis information processing server 3 ′.
The speech synthesizer in the client terminal 1 ′ of this embodiment includes a text data transmission unit 29, a speech unit information / prosodic parameter reception unit 34, a speech unit data reading unit 31, a speech unit connection / deformation unit 35, a speech It has an output unit 33 and a speech unit database DB-1.

FIG. 10 is a flowchart for explaining speech synthesis processing in the client terminal 1 ′ of this embodiment. The details of the speech synthesis process according to this embodiment will be described below with reference to FIG.
Also in the second embodiment, the client terminal 1 ′ transmits the target text data from the text data transmitting unit 29 to the speech synthesis information processing server 3 ′ as in the case of FIG. 6 (S10-1). The speech synthesis information processing server 3 ′ analyzes the sent text data and returns speech segment data and prosodic parameters corresponding to the text data to the client terminal 1 ′. The client terminal 1 ′ receives the speech unit information and prosodic parameters transmitted from the speech synthesis information processing server 3 ′ through the network 2 by the speech unit information / prosodic parameter receiving unit 34 (S10-2).

Next, the content of the processing in the speech unit data reading unit 31 can be executed in the same manner as the processing of the speech unit data reading unit 31 in the first embodiment described above (S10-3).
Next, the speech unit connection / deformation unit 35 sequentially connects the speech units read by the speech unit data reading unit 31 to generate synthesized speech data (S10-4). Here, when speech unit data is connected in temporal order, different speech units are complemented in terms of time or frequency (reference document: Japanese Patent Laid-Open No. 07-072897, “Speech Synthesis Method and Device”). Then, connection is made after signal processing is performed on the speech segment data based on the received prosodic information (S10-4). (Reference: Y. Stylianou, “Applying the Harmonic Plus Noise Model in Concatenative Speech Synthesis.” IEEE TRANSACTIONS ON SPEECH AND AUDIO PROSESSING, VOL.9, NO.1, pp.21-29 JANUARY 2001) (S10-4) .

By doing this, although the processing amount increases somewhat, it becomes possible to control the time length of the speech unit and the fundamental frequency F0 by signal processing compared to the case of simply complementing and connecting, Prosodic accurate synthesized speech can be output, improving the overall quality of the synthesized speech.
Finally, the connected speech segment data is output as synthesized speech at the speech output unit 33 (S10-5).
In the second embodiment, the text data has been described as being sent from the client terminal 1 ′ to the speech synthesis information processing server 3 ′. However, this need not be the case. Text data may be sent to the speech synthesis information processing server 3 ′.

The client terminals 1 and 1 'and the speech synthesis information processing servers 3 and 3' according to the present invention described above install the client terminal program and the speech synthesis information processing program proposed in the present invention in the computer, respectively. It can be realized by executing.
The client terminal program and the speech synthesis processing program proposed in the present invention are described in a computer-readable program language, and are recorded in a computer-readable recording medium such as a magnetic disk or a CD-ROM. The computer is installed from this recording medium or through a communication line. The installed program is decrypted and executed by a CPU or MPU provided in the computer.

  The present invention can be used in fields such as a voice guidance system using a portable terminal, an automatic reservation system, or a voice guidance system in car navigation.

The block diagram for demonstrating the outline | summary of the speech synthesis method by this invention. The block diagram for demonstrating the structure of the whole client terminal used for the speech synthesis method by this invention. The block diagram for demonstrating the structure of the speech synthesis information processing server used for the speech synthesis method by this invention. 4 is a flowchart for explaining an operation of the speech synthesis information processing server shown in FIG. 3. The block diagram for demonstrating the structure of the speech synthesizing means constructed | assembled in the client terminal shown in FIG. The flowchart for demonstrating operation | movement of the speech synthesis means constructed | assembled in the client terminal shown in FIG. The block diagram for demonstrating the structure of the speech synthesis processing server proposed in Example 2 of this invention. The flowchart for demonstrating operation | movement of the speech synthesis information processing server shown in FIG. The block diagram for demonstrating the structure of the client terminal which operate | moves corresponding to the speech synthesis information processing server shown in FIG. The flowchart for demonstrating operation | movement of the client terminal shown in FIG. The figure for demonstrating the outline | summary of the index with which the speech synthesis processing server by this invention was equipped. The figure for demonstrating the outline | summary of the speech unit database with which the client terminal by this invention was equipped.

Explanation of symbols

1, 1 'client terminal 34 speech unit information / prosodic parameter receiving unit 2 network 35 speech unit connection / deformation unit 3, 3' speech synthesis information processing server 40 work memory 9 text data receiving unit 41 MPU
DESCRIPTION OF SYMBOLS 10 Text analysis part 42 Accumulation memory 11 Prosodic parameter acquisition part 43 Data transmission / reception part 12 Speech segment search part 44 Speech output part 13 Speech segment information transmission part 14 Speech segment information / prosodic parameter transmission part
Index-1 speech unit index 29 text data transmitting unit 30 speech unit information receiving unit 31 speech unit data reading unit 32 speech unit connection unit 33 speech output unit
DB-1 speech segment database

Claims (5)

Speech unit information receiving means for receiving speech unit information for identifying speech unit data sent to itself and a prosodic parameter that is a physical parameter for determining the prosody;
Speech unit data reading means for reading out speech unit data from the speech unit database based on the received speech unit information;
Speech unit connection means for generating connection synthesized speech data in order after the read speech unit data is transformed according to the prosodic parameters;
Voice output means for sequentially outputting the generated connection synthesized voice data as synthesized voice;
A client terminal comprising: Text data receiving means for receiving text data sent to itself;
Text analysis means for performing text analysis on received text data and obtaining reading information and prosodic information;
From the prosodic information, prosodic parameter acquisition means for acquiring a prosodic parameter that is a physical parameter that determines the prosody required for speech synthesis;
Speech segment search means for acquiring speech segment information for identifying speech segment data based on the reading information and prosodic parameters using a speech segment index;
Speech unit information transmitting means for adding a prosodic parameter to the speech unit information and sending it to the network;
A speech synthesis information processing server comprising:   A client terminal program that is written in a computer-readable program language and causes the computer to function as at least the client terminal according to claim 1.   A speech synthesis information processing program written in a computer-readable program language and causing the computer to function as at least the speech synthesis information processing server according to claim 2.   A recording medium comprising a computer-readable recording medium, wherein at least one of the client terminal program according to claim 3 and the speech synthesis information processing program according to claim 4 is recorded on the recording medium.

Images