Multiprocessing Japanese Text-to-Speech Engine SMARTTALK Whi

The Datasheet Archive - 100 Million Datasheets from 7500 Manufacturers.

Multiprocessing Japanese Text-to-Speech Engine SMARTTALK Whi

Datasheet Thumbnail

Top Searches for this datasheet

People People Technology
Multiprocessing Japanese Text-to-Speech Engine
SMARTTALK
White Paper
February 2000 Electric Industry Co., Ltd.
1996-2000 Electric Industry Co., Ltd. Right reserved.
SMARTTALK White Paper
People People Technology
Table Contents
SMARTTALK Outline- Features SMARTTALK 2.0-
Supports Prosodic Control Model (PCM) Natural Voice Quality- Enlarged Speech Segment Dictionary- Powerful Dictionary Function Allows Improved Accuracy Simplified User Interface Intelligent Text Analysis Features- Sound Effects Wide Range Applications- Additional Output Data Format
Application Interface
Provided Functions Application Program Structure- List License Information
Specifications
Latest Information
more information about this product, visit Electric Industry homepage http://www.oki.co.jp
SMARTTALK registered trademark Electric Industry Co., Ltd. Japan. other company names products names trademarks registered trademarks their respective companies.
SMARTTALK White Paper
People People Technology
SMARTTALK Outline
SMARTTALK superior Japanese text-to-speech synthesis software package capable supporting multi-thread multi-processing systems connected multiple-line telephone boards such (Computer Telephony Integration). allows various application programs targeted server system incorporate SMARTTALK's powerful Japanese speech synthesis function. When compared previous speech synthesis engines, SMARTTALK superior voice quality, offers enhanced multi-processing multi-thread support capability, offers improved application interface capable utilizing callback functions. Using SMARTTALK DLL, speech synthesis-based systems application programs easily designed highest standards quality performance. SMARTTALK following features. Multi-processing multi-thread support SMARTTALK software engine created servers supporting multi-processing multi-thread activities. simultaneously synthesize different speech patterns active threads processes. been optimized application programs using multiple-line telephone boards such those used systems. SMARTTALK works creating speech synthesis instance each speech synthesis process. using these multiple instances, different speech synthesis performed simultaneously. Speech synthesis, user dictionary, voice attributes thus specified each instance. Natural accurate reading SMARTTALK provides "Prosodic Control Model (PCM)", statistically analyzed model huge natural prosodic database, addition Electric's improved variation "Pitch Synchronous Overlap Add" technology. Thus, natural intonation clear synthesized speech provided. SMARTTALK also supports male female voice types, allowing user choose most appropriate task environment hand. Supporting interface SMARTTALK provides APIs (Application Program Interface) DLLs (Dynamic Link Library), which support Windows programming style. interface designed enhance programming productivity providing simple interface speech synthesis speech output. Because interface supports callback functions, messages such termination speech synthesis vocalization termination, received callback functions within programs. Supports various data output formats Various DATA output formats specified with SMARTTALK including low-level waveform audio device, file, buffer, according situations. select level waveform audio device output device, output voice directly telephony board adapted TAPI (Telephony API) Version above.
SMARTTALK White Paper
People People Technology
Features SMARTTALK
SMARTTALK high quality Japanese text-to-speech synthesis engine capable supporting multi-thread multi-processing speech answer systems equipped with multiple-line telephone boards such CTI. equipped with "Prosodic Control Model (PCM)" statistical speech analysis addition Pitch Synchronous Overlap technology, 110,000-word Japanese dictionary, sound effects additional echo, intelligent text analysis features. various application programs utilize these powerful capabilities output clear natural sounding synthesized Japanese speech. SMARTTALK utilizes same programming interface Version 1.x, applications developed Version Version without modification. features SMARTTALK include following: Supports Prosodic Control Model (PCM) that utilizes huge database provide natural voice quality enlarged speech segment dictionary powerful dictionary function improved accuracy simplified user interface Intelligent text analysis features Sound effects various applications Additional output data format
SMARTTALK White Paper
People People Technology
Supports Prosodic Control Model (PCM) Natural Voice Quality
analyzing large natural speech database, "PCM" statistically improves timing pause reading intonations thus ensuring fluent reading. Adding "PCM" Electric's popular variation "Pitch Synchronous Overlap Add" technology, ensures that improved natural intonation clear synthesized speech provided. SMARTTALK supports male female voice patterns. each provided, allowing user select pattern most suitable environment task hand.
Enlarged Speech Segment Dictionary
previous speech segment dictionary supported only male voice female voice with 22.05 sampling rate, bits bits sampling size. SMARTTALK provides different synthetic speech units. previous version, when 11.025 voice used sampling rate synthesis utilized most telephone lines, voice quality degraded because initially created 22.05kHz speech synthetic unit. Providing 11.025 speech synthetic unit SMARTTALK allows high quality voice synthesis even with sampling rate. addition male female voice patterns, users also specify "Compressed" option addition previous "Decompressed" size. Thus minimum system resource file size reduced about compared with previous size about reducing required resource size load system, number threads that activated same time increased handling capability telephone lines each Windows (Windows 2000) server large system improved. SMARTTALK also provides additional APIs manage enlarged speech segment dictionaries. Refer "API list" further details.
SMARTTALK White Paper
People People Technology
additional speech synthetic units listed below. Types speech synthetic units kHz, bits bits Male Voice Size about 1.15 about 3.85 about 1.17 about 3.86 about 0.84 about 2.71 about 0.76 about 2.40 about 3.92 about 12.82 about 1.54 about 5.27 about 1.54 about 5.27 about 1.11 about 3.70 about 1.00 about 3.30 about 5.19 about 17.55 about 2.80 about 10.43 about 2.95 about 10.43 about 2.04 about 7.12 about 1.86 about 6.47 about 4.84 about 17.55 about 9.65 about 34.45
11.025 kHz, bits bits
Compressed Decompressed Male Voice Compressed Decompressed Female Voice Compressed Decompressed Female Voice Compressed Decompressed Male Voice Female Compressed Decompressed Male Voice Compressed Decompressed Compressed Decompressed Female Voice Compressed Decompressed Female Voice Compressed Decompressed Male Voice Female Compressed Decompressed Male Voice Compressed Decompressed Male Voice Compressed Decompressed Female Voice Compressed Decompressed Female Voice Compressed Decompressed Male Voice Female Compressed Decompressed Male Voice Female Compressed Decompressed Male Voice
22.05 kHz, bits bits
Powerful Dictionary Function Allows Improved Accuracy Simplified User Interface
SMARTTALK DLL's dictionary function been enhanced with addition specialized sub-dictionaries that selected improve reading accuracy enhance usability. Geographical dictionary, biographical dictionary computer dictionary added addition three dictionaries, geographical, biographical, computer (English) dictionaries previous Japanese dictionary results very accurate reading. Geographical dictionary about 3,000 words Biographical dictionary about 2,000 words Computer dictionary (English) about 1,000 words
SMARTTALK White Paper
People People Technology
Supports multiple user dictionaries Some large systems need create maintain user dictionary each function. previous version allows only user dictionary specified time, SMARTTALK allows multiple user dictionaries selected simultaneously. Because user maintain user dictionary each file, word registration maintenance proceed much more smoothly. Provides file import function sample application addition previous tools registering words, SMARTTALK provides file import function sample application program. Using this application program, list formatted user dictionaries made Microsoft Excel, etc. directly imported SMARTTALK user dictionaries.
Intelligent Text Analysis Features
Abbreviated numerical formats utilized wide variety abbreviated day, date time, phone number formats available use. Examples: 99/06/12 Read "Kyu Roku Gatsu Nichi" 6:12 Read "Roku Fun" Alphabetical words reading mode alphabetical words reading mode specified follows; Roman alphabet character reading, English Reading, Auto detection. Therefore, words that must read Roman letters, such mail address read correctly. [Auto detection] [Alphabetical reading [Alphabetical reading Read known word using registered contents from dictionary Read unknown word with rules Read known word using registered contents from dictionary Read unknown word alphabetically Read both known unknown word alphabetically
SMARTTALK White Paper
People People Technology
Sound Effects Wide Range Applications
SMARTTALK provides various sound effects functions addition previous plain text-to-speech. extensively used game application programs entertainment systems. using sound effects function, following four different types sound effects added synthesized voice. None High frequencies emphasis Echo Default frequencies listened well fields. echo effects speech output (further options specified). Five echo settings, "Bathroom", "Cave", "Hall", "Dome" "Echo" specified added voice. Output mechanical synthesized voice such robot sound.
Robot
each echo setting, following parameters specified. Initial delay Echo delay Echo repeats 1,000 1,000 times
followings some examples changeable echo effects specifying parameters. Example Bath room (default) Cave Hall Dome Echo Initial delay (ms) 1000 Echo delay (ms) 1000 Echo repeats (times)
Additional Output Data Format
SMARTTALK DLL, users specify output format from level wave form audio device, file, buffer their specific purposes. addition previous data format, SMARTTALK provides additional output formats "A-Law" "Mu-Law" buffer types. These formats allow user directly output synthesized voice telephony board that supporting data format such modem.
SMARTTALK White Paper
People People Technology
Application Interface
Provided Functions
SMARTTALK individually synthesize different speech each instance. instance process unit SMARTTALK DLL. Each instance provides following functions. Text-to-speech synthesis output synthesized from text data input. output destination specified device, file buffer. output format supports follows: Sampling rates 11.025 22.05 Sampling size bits bits bits bits bits bits Output data format A-Law Mu-Law A-Law Mu-Law A-Law Mu-Law
Intermediate language speech synthesis output speech synthesized from intermediate language input. output destination specified device, file buffer. Performing speech synthesis utilizing intermediate language that describes phonetic prosodic information efficient known sentences. Because contents read known, after text conversion intermediate language, prosodic symbols edited speech synthesis performed. This method provides more natural speech than straight text-to-speech synthesis. Attributes settings following attributes specified about voice. Output format bits 16bits, A-Law Mu-Law kHz, bits bits, A-Law Mu-Law kHz, bits bits, A-Law Mu-Law Number reading mode Auto detection, read numerical value, read separate numbers, Terminators Comma, Period, Touten (Japanese comma), Kuten (Japanese period), successive Voice volume levels Voice pitch levels Speech speed levels Voice type Male Female Intonation levels Vowel devoicing Symbols read Specify range shift code Pause interval between sentences levels Voice quality conversion levels Sound effects types
Device output control
Pause, Resume, Stop, Skip
User dictionary maintenance
Entry word, Delete word, word information
SMARTTALK White Paper
People People Technology
SMARTTALK White Paper
People People Technology
Control commands Embedding control commands into text intermediate language describing phonetic prosodic information control reading. commands allow pause, comment out, mark, chime/buzzer sound output, musical scale output settings, well attributes settings.
Others Unknown word search, splitting into syllables, reading time prediction, intermediate language generation, error checking intermediate language.
Application Program Structure
following chart shows process flow simple application program using Japanese text-to-speech synthesis SMARTTALK DLL. Each rectangle this chart refers function call SMARTTALK DLL.
SMARTTALK White Paper
People People Technology
Start
Initialize
Create speech synthesis instance
attributes instances
Load user dictionary
Text speech synthesis
Call callback functions
Delete speech synthesis instance
Terminate process
Exit
Figure process flow chart application program
speech synthesis instance created each thread process user application program. following types speech synthesis instances. Users should create speech synthesis instance applicable their required task. Text-to-speech synthesis instance Intermediate language speech synthesis instance Intermediate language that describes phonetic prosodic information generation instance
Each created instance runs individually. attributes (speech speed, voice type, etc.) user dictionaries must specified each instance. attributes specified, default values used. Default values specified individually each system installed, Windows (Windows 2000) used, they specified each user account.
SMARTTALK White Paper
People People Technology
course, different speech synthesis performed simultaneously multiple instances.
SMARTTALK White Paper
People People Technology
With exception device control (Refer chapter "API list"), control each instance executed same thread running upon. Application program Threads
Text-to-speech synthesis instance
Intermediate language speech synthesis instance
Intermediate language generation instance
User dictionary
SMARTTALK
System dictionary
Figure Relationship between application programs instances
text-to-speech synthesis process, three following output destinations specified: Audio device Buffer File audio device specified device simply output speech through soundboard system, specify WAVE_MAPPER device available soundboard will selected automatically. audio device output, device must supported Microsoft's low-level waveform audio device. output speech through board with proprietary interface, output destination text-to-speech synthesis specified either buffer file. After registering appropriate callback function, output text board sentences callback function. text-to-speech synthesis functions provided SMARTTALK execute processes synchronously. other words, function does return until speech synthesis data input buffer have been output audio device, buffer, file) output been terminated. This means application cannot perform other during text-tospeech synthesis processing. SMARTTALK does however, allow application programs register callback functions perform calls pre-defined timings during speech synthesis process. create more interactive application programs, register your callback functions, which perform interactive tasks. following shows timings calling callback function.
start speech data generation sentence
SMARTTALK White Paper
People People Technology
speech data generation sentence start speech data output sentence speech data output sentence When recoverable error issued When speech data generation process reaches marked point input buffer
sentence refers speech data generation unit that SMARTTALK recognizes sentence delimiter (terminator) input text data. speech synthesis process, text data passed input converted into speech data output destination, each speech data generation unit. speech data generation unit specified sentence terminator attribute. default, sentence terminator attribute Kuten (Japanese period). example, sentences fine today. will rain tomorrow. given input, fine today" will rain tomorrow" each recognized sentence speech data generated both. SMARTTALK automatically insert punctuation into sentence determines that long. callback functions receive instance handle, start byte position sentence input buffer, start byte position next sentence, generated speech data size, optional data application program
List
following shows list. Function name Initialization termination ttsInitiate ttsTerminate Instances management ttsCreateTextSyn ttsCreateIMLSyn ttsCreateIMLGen ttsDeleteInstance ttsCreateTextSynEx ttsCreateIMLSynEx Description
Initialize Terminate
ttsCreateIMLGenEx
Create text-to-speech synthesis instance Create intermediate language speech synthesis instance Create intermediate language generation instance Delete instance Create text-to-speech synthesis instance (extended) Create intermediate language speech synthesis instance (extended) Create intermediate language generation instance (extended)
Speech synthesis ttsTextToDevSyn ttsTextToBufSyn ttsTextToFileSyn ttsIMLToDevSyn ttsIMLToBufSyn ttsIMLToFileSyn ttsCheckIML
Synthesis speech from text output device Synthesis speech from text output buffer Synthesis speech from text output file Synthesize speech from intermediate language that describes phonetic prosodic information output device Synthesize speech from intermediate language that describes phonetic prosodic information output buffer Synthesize speech from intermediate language that describes phonetic prosodic information output file Check syntax intermediate language that describes phonetic prosodic information
SMARTTALK White Paper
People People Technology
ttsMemoryFree ttsRegCallback ttsUnRegCallback ttsLoadUserDic ttsUnloadUserDic ttsTextToIMLSyn
Attribute
ttsSetWaveFormat ttsGetWaveFormat
Clear output buffer Register callback function Unregister callback function Load user dictionary Unload user dictionary Generate intermediate language that describes phonetic prosodic information output data format output data format settings
ttsSetPCMFormat ttsGetPCMFormat ttsSetNumRead ttsGetNumRead ttsSetTerminator ttsGetTerminator ttsSetVolume ttsGetVolume ttsSetPitch ttsGetPitch ttsSetSpeed ttsGetSpeed ttsSetVoiceType ttsGetVoiceType ttsSetIntonation ttsGetIntonation ttsSetVoiceless ttsGetVoiceless ttsSetSymbolRange ttsGetSymbolRange ttsSetSymbol ttsGetSymbol ttsSetInterval ttsGetInterval ttsSetEffect ttsGetEffect ttsSetEffectEx ttsGetEffectEx ttsSetVoiceTypeEx ttsGetVoiceTypeEx ttsSetVoiceQuarity ttsGetVoiceQuarity ttsSetTimeFilter ttsGetTimeFilter ttsSetDateFilter ttsGetDateFilter ttsSetTelFilter ttsGetTelFilter ttsSetMailDateFilter
output data type output data type settings number reading mode number reading mode settings terminator terminator settings voice volume voice volume settings voice pitch voice pitch settings speech speed speech speed settings voice type voice type settings voice intonation voice intonation settings vowel devoicing vowel devoicing settings range symbols read range symbols read settings symbol reading symbol reading settings interval between sentences interval between sentences settings sound effects sound effects settings sound effects options (Echo) sound effects option settings (Echo) voice type(Extend) voice type settings(Extend) voice quality voice quality settings text analysis abbreviated time text analysis abbreviated time settings text analysis abbreviated date text analysis abbreviated date settings text analysis abbreviated telephone number text analysis abbreviated telephone number settings text analysis abbreviated mail date
SMARTTALK White Paper
People People Technology
ttsGetMailDateFilter ttsSetDayFilter ttsGetDayFilter ttsSetAlphaRead ttsSetAlphaRead ttsSpeakerInfo
text analysis abbreviated mail date settings text analysis abbreviated week text analysis abbreviated week settings alphabets reading mode alphabets reading mode settings speech segment dictionary information
Device output control ttsPauseTalk ttsResumeTalk ttsStopTalk ttsSkipTalk User dictionary management ttsOpenUserDic ttsCloseUserDic ttsPutEntry ttsDelEntry ttsGetInfoEntry ttsGetListEntry Others ttsUnknownWord ttsGetSyllable ttsExpectTime ttsGetIMLWord added
Pause output device Resume output device Stop output device Skip output device
Open user dictionary Close user dictionary Entry word user dictionary Delete word from user dictionary word information from user dictionary word list user dictionary
Search unknown word Split into syllables Expect reading time Generate intermediate language from specified word
License Information
number speech syntheses that performed same time depends number licenses. This number licenses refers number available speech synthesis instances that simultaneously CPU. This number independent number processes threads. following shows examples eight user licenses. this example, there process consisting eight threads, each which generates instance, running system.
Threads Speech Speech Speech Speech SynthesisSynthesis Synthesis Synthesis Instance Instance Instance Instance Speech Speech Speech Speech SynthesisSynthesis Synthesis Synthesis Instance Instance InstanceInstance System
Application program
SMARTTALK White Paper
People People Technology
this example, there processes, each which generates four instances, running system. Application program Threads
Speech Speech Speech Speech synthesis synthesis synthesis synthesis instance instance instance Instance Application program
Threads
Speech Speech Speech Speech synthesis synthesis synthesis synthesis instance instance instance Instance System
SMARTTALK White Paper
People People Technology
Application
Speech synthesis instance
Application
Speech synthesis instance
There eight application programs, each which generates speech synthesis instance, running system.
Application
Speech synthesis instance
Application
Speech synthesis instance
Application
Speech synthesis instance
Application
Speech synthesis instance
Application
Speech synthesis instance
Application System
Speech synthesis instance
Note When instance created expends license regardless whether performing speech synthesis. Note application programs, example, number telephone lines generally equals number licenses. server controls eight telephone lines, number user licenses eight.)
incorporate SMARTTALK into product with intention reselling otherwise redistributing third party, sublicenses must obtained beforehand Electric Industry Co., Ltd.
SMARTTALK White Paper
People People Technology
Specifications
Contents Package Japanese text-to-speech synthesis Sample application programs Source files Executable modules Makefile User dictionary edit tool User's guide (document) reference (online help) This package includes free answering (for three accounts). Standard price (Exclude tax) Product name Price (exclude tax) threads 198,000 threads 300,000 Additional answering accounts) 37,500 incorporate SMARTTALK into product with intention reselling otherwise redistributing third party, sublicenses must obtained beforehand Electric Industry Co., Ltd. Function Specification Function Output format Description level waveform audio device Wave file Synthesized speech buffer PCM/A-Law Mu-Law kHz, 11.025 kHz, 22.05 kHz, Japanese text written Kanji Kana characters (shift JIS) Intermediate language that describes phonetic prosodic information About 110,000 words Pitch Synchronous Overlap Male Female
Synthesized speech format
Input format
Dictionary Synthesis method Synthesis unit Voice type Operating environment Computer
Disk space Media
single processor computer running Japanese Windows95 Japanese Windows Japanese Windows Version Japanese Windows 2000 speech output device, must capable supporting level waveform audio service. Standard: Maximum: CD-ROM
SMARTTALK White Paper
People People Technology
Development Environment
SMARTTALK developed with Japanese Visual (Service Pack Japanese Windows (Service Pack create application programs using SMARTTALK DLL, should prepare developing environment separately (Japanese Visual C++5.0 recommended) SMARTTALK package does include programming language environment required develop application programs.

Multiprocessing Japanese Text-to-Speech Engine SMARTTALK Whi

Top Searches for this datasheet

Other recent searches