Hong Kong's Cyberworkshop developed the speech recognition module for use in the Name Dialer.

The market for speech recognition products has been experiencing rapid growth, particularly in the last two years. Continued improvements on technology and product accuracy have contributed to shorter time to market, further stimulating product development and resulting in more new products. Most automatic speech recognition (ASR) systems in Hong Kong, South Korea and Taiwan are supplied by total solution providers who integrate the technology into server-based, computer telephony integrated (CTI) products.

Most of the available solutions are incorporated into software or hardware systems. The applications continue to expand for ASR systems, and now include call centers, handheld devices, toys, security systems, even in-vehicle devices. Makers note fast-paced growth in the handheld device marketplace, although they agree that server-based applications will remain strong in the coming years.

Most of the 15 suppliers of interactive voice response (IVR) system in Hong Kong are also the enablers of ASR technology. As total solution providers, these makers also supply imported hardware equipment for setting up an IVR system. Many IVR boards, such as those from Dialogic, support seamless integration of software-based speech recognition software from leading speech technology vendors such as Nuance. Using these tools, suppliers create customized ASR solutions that usually entail authentication and message forwarding for their customers.

In Taiwan, currently available ASR systems are designed for targeted applications. Makers see the growth potential of ASR technologies in the region, and forecast even broader adoption in the coming years. They attribute the growth to the increasing use of voice command technologies in more mobile phones and toys. Voice dialog technologies are also applied to specialty fields such as auto attendant systems and weather inquiry systems. In addition, voice input is gaining traction in GPS, portable multimedia devices and smart phones.

Taiwan has less than 10 makers of ASR technologies. The active players include Applied Speech Technologies (AST), CWeb Technology Inc., Cyberon Co. and Penpower Technology Ltd.

Currently, there are three major ASR product makers in South Korea, mostly providing engines to system integrators of CTI system. Immature technology and the language barrier inhibit market growth, but makers are rather optimistic that a market rebound is in the offing and will continue through 2010. Makers forecast that the existing UI will eventually evolve from graphic user interface (GUI) to voice user interface (VUI).

ASRs from South Korea were previously used in stock markets, traffic services, weather services and airline information services. Applications have now expanded to different industries, such as finance, manufacturing and communication. Among the active makers of ASR systems in the region are PentaxVoiceware Co. Ltd, HCI Lab Co. Ltd, YESpeech Co. Ltd and KMIC Global Inc.
ASR systems integrated into various voice solutions
Server-based and toy applications upcoming
In-vehicle navigation, security applications promising
Emphasis on ODM, domestic markets

	South Korea's PentaxVoiceware introduced an ASR engine, the VoiceEZ, which supports 97 percent deployment accuracy and realtime response.

Cyberon specializes in speech recognition and text-tospeech (TTS) technologies for embedded systems. The company offers speech interface solutions for handheld devices such as mobile phones and pocket PCs.

The company's solutions are already ported on Java and commonly used platforms from Texas Instruments (TI), ADI, Agere, Infineon-APOXI, MediaTek and Sunplus. It has also supported many mainstream smart phone platforms, such as Symbian UIQ, Symbian series 60, Microsoft Smartphone, Microsoft PPC Phone and Linux OS. Cyberon's voice solution works with Bluetooth headsets to enable direct access of speech recognition application.

Penpower focuses on integrating core technologies into information appliances (IAs), wireless communication products, smart home appliances and Internet. The company has provided embedded solutions for various platforms. It invests 10 percent to 16 percent of annual sales in R&D, and employs 39 R&D engineers.

AST provides project-based speech recognition implementations. The company has provided some speech recognition engines in Chinese. The company's speech recognition technologies are used in CTI, robots, telematics, security systems and multimedia devices.

Among Hong Kong suppliers that can integrate ASR systems into their total solutions are Expert Systems IVR (Asia) Co. Ltd (ESI) and KanHan Technologies Ltd.

ESI, which focuses on IVR applications, enables voice automation systems by utilizing speech recognition and TSS technology from Nuance. The company has over a decade of experience serving telecom companies. Aside from Nuance, ESI has close partnerships with Intel, Envox, Avaya, Adlink, UniVoice, System Solutions, Vertical, AudioCodes, Stratus Technologies and mainland China's EVOC.

Unlike other suppliers, Cyberworkshop Ltd's total solutions end up in consumer devices. Strategically located near the City University of Hong Kong, the company specializes in providing hardware and software speech recognition solutions. Its services range from designing software and circuit layout, prototyping, voice recording and editing, test jig designing to assembly of finished modules.

KanHan uses ASR as a vital component in some of its IVR deployments. The company offers voice recognition for integration into its HanPhone and HanVoice Server software for transforming Websites into IVR systems. At present, KanHan is developing a voice recognition-enabled IVR system for Hong Kong's Public Housing Authority. The project will enable the agency's clients to inquire about payment-related information. It can identify a caller by voice and give payment details in both Cantonese andEnglish.

With the ASR system considered as a component of the IVR solution, suppliers such as ESI and KanHan base their prices on the complexity of the project and number of ports. For Cyberworkshop, the main price determinant is the complexity. The company does not have a standard solution. Each solution is custom-made.

	Penpower of Taiwan has launched the Voicewriter, which features continuous voice input and sentence voice input.

Suppliers expect the telecom market to continue to be dominated by server-based speech recognition systems.

Device developers such as Hong Kong's Cyberworkshop do not see this trend changing. Cyberworkshop has a few telephony applications. Its Name Dialer, for instance, allows users to place a telephone call automatically by just saying the name of the person to be called. The modular device is placed on the desktop, hooked up to the telephone and records names and telephone numbers. It can store 50 names, with optional caller ID.

Currently, toys comprise the main application for speech recognition technology as applied in consumer products. The next sphere is the appliance market.

One of Cyberworkshop's most advanced products is an educational toy that can recognize more than 100 sets of words. It uses Hidden Markov Model (HMM) technology, the same technology used in server-based systems. According to Cyberworkshop, HMM is a proven technology that has been used in speech recognition for many years. It was developed by Dr. Lee Kai-fu, who now heads Google's R&D center in mainland China.

The company focuses on three main areas: speech recognition, long-speech products such as audio books, and remote controls for toys. Cyberworkshop caters to about 20 major clients annually, including an educational toy brand, Vtech.

Cyberworkshop accepts custom-design services that may range from four to eight weeks for short projects, and three months for developing a new module. Among the many products that utilize Cyberworkshop's solutions are electronic learning devices, talking clocks and books, voice-controlled radio, TV games, and PC accessories. Cyberworkshop designs the hardware, circuitry and software. The manufacturer usually makes some modifications as they have to take into account the production costs, before the design is finalized.

Cyberworkshop uses ICs from Sensory Inc., which is also its main technology partner. The company is also a distributor of sensory ICs in greater China.

Another potential future application for ASR systems is in vehicles for controlling navigation and car audio systems. In Taiwan, Cyberon's voice-recognition software engine is designed specifically for mobile devices. It supports speakerindependent speech recognition. Cyberon's speaker-independent technology assists users to search a contact or call any person in his or her contact list. In addition, it can also help to access games, e-mail, calendar or the Web over the phone.

Penpower has five kernel technologies: handwriting recognition, speech recognition, OCR, biometrics and network security. The company has provided embedded solutions on various platforms. Penpower's own-brand, Voking, is a Windows CE-based voice command/control application. It features voice-activated dialing, voice searching, voice music playing and voice command/control.

AST works with manufacturers and developers to incorporate the company's speech recognition technologies into their mobile phones, PDAs, business intelligence (BI) software, security devices and in-vehicle two-way radios.

South Korea's PentaxVoiceware offers ASR solutions for call centers of server markets, and language education for the PC and Internet. It also supplies products for embedding into telematics, mobile phones, robots and home automation. The company has completed the commercialization of Korean and English versions of its ASR systems, while boosting the Chinese and Japanese versions.

HCI Lab has released the PowerASR, which is speakerindependent, and supports broad vocabulary and flexible recognition. The model features a speech recognition system for isolated, continuous speech and connected digit recognition. It features N-Best decoding, confidence scores at sentence, and word level tunable voice diction sensitivity. The PowerASR is available in four languages: Korean, English, Chinese and Japanese.

KMIC Global provides hybrid engines that integrate IVR and ASR solutions for personal telematics and location-based services (LBS). The service supports voice data center, which is a network-based hybrid speech recognition technology. The service enables drivers to find directions to a destination ask for emergency help and receive traffic alerts.

ASR system developers in Taiwan, Hong Kong and South Korea focus respectively on their own domestic markets. Taiwan's speech recognition suppliers, which pursue ODM and own-brand business, focus more on the domestic market, as they mainly offer Chinese solutions.

AST and Cyberon, which mainly provide ODM business, generate about 95 percent of sales from the domestic market. AST's overseas business covers mostly the Middle East. The company's major domestic clients include Taiwan's National Security Bureau, AsusTek and Gigabyte.

Cyberon's speech recognition solutions have been used in more than 60 models of mobile phones, spread over a combined 8 million mobile phones and pocket PCs. In Q1 2006, the company's solutions were embedded in 1.5 million units of handheld devices. Most of Cyberon's sales come from the domestic market. Its major mobile phone clients include HTC, Inventec Appliances, Motorola, BenQ, AsusTek and Eten.

Penpower sold 241,350 units of handwriting and speech recognition products in 2005. The company also offers projectbased speech recognition implementation services to its ODM clients. About 91 percent of Penpower's sales come from own-brand manufacturing (OBM). The rest is generated from ODM business. The overseas region accounts for 66 percent of Penpower's sales, and 34 percent comes from the domestic market. The company's licensed clients include Garmin, Maction Technologies and KBT Mobile.

Aside from the local market, ESI also has customers in Malaysia, Thailand, the Philippines, Taiwan and mainland China. Ninety-five percent of KanHan's customers are local companies. Although 90 percent of Cyberworkshop's customers are Hong Kong-based manufacturers of toys and electronic products, many of them are branch offices with headquarters in the United States, Europe and other regions. Ten percent are overseas manufacturers. North America, Europe and East Asia are the major export markets. About 80 percent of PentaxVoiceware's ASR solutions are distributed in the domestic market.