About VORERO - VORERO

HOME > Voice recognition middleware > About VORERO

What does VORERO work for? VORERO is the middleware for voice recognition, which can use embedded system for various scene. Development of voice recognition technology for use as an interface for computers and embedded systems has progressed for several years, and is presently employed in car navigation systems, cellular phones, and robotic pets. As the advance of information technology continues, we believe the voice recognition technology will become common not only in automotive and mobile electronics applications, but also in appliances and electronic products for the home. Voice recognition technology also promises to enhance barrier-free usability of appliances and IT equipment for people with physical handicaps, and to help eliminate the digital divide by enabling the utilization of information-age tools without the need for computer literacy. Recognizing the potential benefits of voice recognition technology, Asahi Kasei’s development effort has focused on optimum embeddability, high performance, and low cost. The name VORERO is taken from “voice recognition that is robust” due to its outstanding performance in the midst of the real-life noise of office, automobile, and home environments.
go to top

Features Speech recognition middle ware based on advanced acoustic analysis and HMM (Hidden Markov model ) matching technology. Speaker-independent voice recognition using proprietary phoneme model. Easy configuration of vocabulary for recognition by alphabetic input. High performance Outstanding robustness to noise in real life environments. High tolerance to noise. Not limited to headset microphones. Outstanding keyword spotting and rejection, making voice interface easy to achieve with continuous utterances using near-natural language. Ignores unnecessary surrounding words and out-of-vocabulary words. Low cost Low processor load and low memory requirement enable low system cost. Acoustic analysis Feature extraction method to achieve robustness to noise. Noise canceller Removes additive noise from unknown sources, such as constant background noise. Frequency equalizer Compensates for frequency distortion due to microphone, distance from mouth to microphone, and indivisual voice characteristics. Speaker sound canceller Removes additive noise from known sources, such as sound from TV and PC speakers.-with Hands free middleware VOCLE. Ability to combine speaker-independence and speaker-dependence Ability to combine word model and phoneme model Support for multiple languages North American English North American Spanish Canadian French Japanese Korean Mandarin Cantonese British English French German Spanish Dutch Portuguese Italian Swedish Russian
go to top

Architecture VORERO engine VORERO data Acoustic model Vocabulary Netowork (Dictionary)

VORERO engine Voice input from a microphone is converted to a digital signal by PCM. The VORERO engine performs preprocessing for removal of noise, acoustic analysis, and Viterbi matching. VORERO data Using a unique architecture, VORERO may be optimized for various applications by modifying the acoustic model and recognizable Vocabulary Network. Users may freely prepare Vocabulary Networks with word orders and vocabularies for recognition. Acoustic model The acoustic model used to perform Viterbi matching may be configured for dynamic switching between phoneme and word models, mixed recognition of registered speaker-dependent word models and prepared speaker-independent word models, and mixed recognition of Japanese and English. Functions to optimize the acoustic model for specific applications are also available, including optimization for ambient noise encountered during cellular phone use and for road noise encountered in automotive use. Japanese phoneme model is included as standard in the Japanese version of VORERO, while the English phoneme model is included as standard in the English version. Word models are included per user requirements, and in some cases are subject to option fees. Upon user request, Asahi Kasei will also perform word model production for a fee. VORERO Vocabulary Network (Dictionary) Prior to executing voice recognition with VORERO it is necessary to configure the vocabulary and word order for recognition. This is called the Vocabulary Network, and is incorporated into the voice recognition engine as binary image data. A Vocabulary Network production tool (VORERO SDK) is available from Asahi Kasei. VORERO's unique functions for keyword spotting and rejection of unnecessary or out-of-vocabulary words also depend on the configuration of the Vocabulary Network. More complex recognition requires a more complex Vocabulary Network. Standard macros for difficult to produce Vocabulary Networks are available from Asahi Kasei.
go to top

Platforms Windows Windows NT/2000, XP, CE (VORERO available as a Visual C++ 6.0 library) Embedded Micro-controllers : ARM7, ARM9, StrongARM, XScale, SH2, SH3, SH4, TX, Vx, VR, M32R DSPs: TMS320C5000, TMS320C6000 VORERO is not dependent on a specific RTOS, and has been used with Nucleus Plus on ARM, NORTi, VxWorks, Windows Automotive, Windows CE, PocketPC, Windows CE.NET, Linux and DSP/BIOS Number of MIPS and amount of memory required for the VORERO engine will vary depending on the number of words in the vocabulary for recognition and the complexity of the Vocabulary Network. VORERO is scaleable from tens of MIPS to hundreds of MIPS.
go to top

Product category VORERO middleware is available in three product categories: VORERO-11 [11.025kHz sampling] Devices Automotive accessories(car navigation systems and audio systems) PCs PDA Home appliances and electronics Karaoke systems Robots Toys Educational equipment etc... VORERO-8 [8kHz sampling] Devices Cellular phone Mobile communications devices etc...
go to top