SinoVoice’s Application on Mobile Phones--part 1

[alexstary]

1. Characteristics of SinoVoice’s three core technologies

1.1 Characteristics of SinoVoice’s Handwriting Recognition Technology — eJHWR
1.1.1 System Structure
On-line handwriting recognition is the process of transferring orderly trail information, which is formed while the user writs on a writing device, into Chinese in-code. It actually is a mapping process from coordinate sequence of writing trails to Chinese in-code.

1.1.2 Technical features
SinoVoice’s HWR technology deploys multi-core integration technology which is based on sentence structure self-learning algorithm and characteristic statistic algorithm, having the characteristics as follows:
High recognition rate
High recognition speed
No restriction on writing orders
The size of data library can be extended or conpressed

Features of eJHWR
Index Indicator
Character set Support GBK，BIG5，Unicode
Recognition sphere Chinese（simplifies Chinese 6763 + traditional
Chinese 5401+ GBK 21003）numbers、English
symbols、strokes
Recognition accuracy of standard character 99.9%
Recognition accuracy of casual character 99%
Recognition accuracy of characters in casual stroke orders 96.8%
Recognition speed 0.4 second/character - 0.6 second/ character（under MIPS 33M）
Dictionary size 600K-1.5M
Space taken by program 40-80K
Stack space < 4K
Self adaptive capability support
OS WinCE、Embedded Linux、Symbian、Nucleus、
Palm OS、HOPEN、pSOS、UCOS etc.
CPU ARM9、ARM7、StrongARM、MIPS、DragonBall
（Motorola 68000)EZ（16M）and VZ（33M）、
SH3、SH4、Xscale etc.

1.2 Technical Characteristics of SinoVoice embedded TTS product –eJTTS
1.2.1 System Structure

TTS is also called Text-to-Speech，which is a technology to transfer any combination of text files into audio files by computer and plays the voice by sound card, telephone card and other multimedia apparatus. Simply put it, TTS can make the machine read out the text files.

1.2.2 Technical Features:
SinoVoice’s TTS is a core technology formed by ‘adding voice library compression’ algorithm and ‘voice tone changing’ algorithm, which enables it to reserve the excellent synthesis effect to the maximum extent, with the characteristics as follows:
 the system outputs voice in high clarity
 the voice output by the system is easily understandable
 the naturalness of the voice is almost the same as people’s voice
 the size of voice library can expand or contract, which ranges from 400K-11M，some specific voice library is less than 100K
Index Indicator
Language Chinese
Chinese English mixed recognition Support
Voice Female Mandarin, male Mandarin,
female Taiwanese, female Cantonese
Voice library Around 400K-11M
Other occupied space Vocabulary library：200-500K
Program： 50-100K
Extra needed space while operated： 16K
Occupied space of CPU (the software includes special voice library, customized voice library, program code, etc.) Size of software package（Take voice library of 1.1M as example） Correspondent CPU resources
1.8M 35～40 mips
2.5M 5～6 mips
4.8M 1mips or so
Audio output format WAV 8K 16bit（if needed can transfer into other formats）
Accuracy 96.59%
Clarity 94.00%
Degree of understandability 97.83%
Naturalness 4.09（single sentence），4.26（paragraph）
Character set Support GB_2312，GBK，BIG5
OS WinCE、Embedded Linux、Symbian、Nucleus、Palm OS、HOPEN、pSOS、UCOS etc.
CPU ARM9、ARM7、StrongARM、MIPS、DragonBall（Motorola 68000)EZ（16M） and VZ（33M）、SH3、SH4、Xscale, etc

1.3 Technical characteristics Of SinoVoice’s embedded Automatic Speech Recognition – eJASR

1.3.1 System structure

Automatic Speech Recognition is the technology that can makes machine recognizing and understanding audio information and convert it into correspondent files or command. In the end of 2001, SinoVoice developed the automatic speech recognition technology: eJASR, which is in the technology category of ‘command recognition and control’. In year 2002, SinoVoice transplanted this technology into Pocket PC.

1.3.2 Technical characteristics
 Index Indicator
Language Mandarin
Recognized content Words
Size of model 300K
Vocabulary size 6K/200 words
Size of program < 40K
Extra space taken while operated 300K
Audio input format WAV 8K 16bit（if needed can transform into other formats）
Accuracy 96%（less of 200 words）
OS WinCE、Embedded Linux、Symbian、Nucleus、Palm OS、HOPEN、pSOS、UCOS etc.
CPU ARM9、ARM7、StrongARM、MIPS、DragonBall（Motorola 68000)EZ（16M）和VZ（33M）、SH3、SH4、Xscale etc.

2. SinoVoice’s software products for mobile phones
2.1 The application software of eJHWR on mobile phones
2.1.1 Handwriting input software

Handwriting Input software which deploys SinoVoice’s handwriting recognition technology –jPEN, is a software developed for mobile phone to realize handwriting input function for its short message service, contact book, notebook, and other functions.

Input text for short message, contact book, notebook, schedule and other functions by handwriting in stead of by frequent typing keys on keyboard, and avoid frequent switching of input methods when characters are mixed with symbols. Even when confronted with characters of uncertain pronunciation, input can still be made easily.

In the following picture：The input interface of contact book in Windows CE OS

Handwriting input method provides optional characters, related characters and other frequently used functional keys, makes the mix input of Chinese, English words and symbols very convenient.

Rapid Location
The handwriting pen can be used as the mouse to locate rapidly, which is more flexible than operating on keyboard

Rapid search
When the menu has too many levels or the user doesn’t know the exact category of the searched information, handwriting input can cut short the aimless keyboard searching and get the result rapidly

Painting
Use the handwriting pen as a drawing brush, paint any pictures you like, add more fun to your life.

Surf on Internet
Use the handwriting method to chat on-line, input email content and search information on Internet, which is convenient and fast.

2.2 Software products of eJTTS on Mobile Phones
2.2.1 Short Message Reader

Short Message Reader deploys SinoVoice’s embedded Text-to-Speech technology –eJTTS. It’s software providing short message audio broadcast function for mobile phones.

Pre-set for Short Message Reader

After the premium set, Short-message reader will be in a state of readiness all the time until the mobile phone is turned off or turning off the short message reader.

Audio notification of incoming short message
When a short message comes, the mobile phone will say automatically ‘Beep, X o’clock, XX minute, receive one (or several) short message(s)’

After receiving the short message and informing the user, it will read the short message if the user press any key within 3 seconds

In the process of reading, the user can stop it by pressing any key

The above is the circumstance of automatically receiving and reading short messages. If choose ‘short message reader’ positively, the ‘short message reader’ will provide the following information in order:
☆ How many short messages have not been read
☆ Read the un-read short messages in batch

2.2.2 Mobile Reading Power

Mobile Reading Power is a software that deploys SinoVoice’s embedded TTS technology – eJTTS, developed for mobile phones to provide audio output function, including 4 functions of broadcasting incoming calls, broadcasting time, broadcasting schedule and broadcasting statue.

Broadcasting incoming calls
Incoming call broadcasting is an excellent attached function that has huge attraction to mobile phone users. It is like the software of QQ that when a message comes, it will cough lightly like a polite secretary reminds the user:

When the phone number is already recorded:
‘ Beep, someone is calling, shall we answer it?’

When the phone number has not been recorded:
“Beep，010 82826886 is calling, shall we answer it？”

After setting with SinoVoice’s technology, it can judge the calling zone
“Beep——， heading with 82，this is a phone call from HaiDian district”
“Beep——， a international long-distance call from America..…”

Voice time reader
Voice time reader turns the mobile phone into a speaking watch. By pressing a certain key, the mobile will speak ‘ Beijing time 11: 29 AM，May 29th，Thursday’

Schedule voice reminder
The schedule reminder is most amiable and very kind to mobile manufacturers and to the users.
 “schedule voice reminder, press any key to continue ——”，if there is no response, it stops automatically.

After the user press a key
‘1，Today, the representative of Guang Zhou branch will come back to Beijing’. 2，Today, we should quote for Sim Phone Company.’
‘ Tomorrow，there is a developing zone exportation promotion conference for software’
 If the user double clicks the key (or press a certain key), it will stop playing immediately.

Status reminder
The state reminder includes the following functions:

 Voice remind the battery status, if the battery needs to be recharged, the system will reminder automatically:
‘The power is low. Need to recharge the battery.’
 Balance reminder: If the balance is low, the system will reminder the user automatically:
‘The balance is low. The phone needs to be recharged.