Hong Kong's Applied Science and Technology Research Institute (Astri) has unveiled its first project, a voice-recognition engine, which it hopes to commercialise by licensing to publishers of English-learning software and Internet portals.
The software, which Astri has dubbed Sonic Ranger, has been in development for just under a year and was beta tested on about 90 students in a language-study camp this summer. Based on research done at Polytechnic University and Hong Kong University of Science and Technology, it is one of a number of projects under way at the HK$3 billion government-funded organisation.
Astri vice-president for information technology Roland Chin Tai-hong said talks were ongoing with companies interested in licensing Sonic Ranger for use on compact discs and in online learning courses. Interested parties include groups that provide educational materials to schools in Japan and mainland China.
The engine breaks sounds into component parts - or phonemes - and compares what the user says into a microphone with pre-recorded sounds collected from about 500 young speakers. This allowed for oral practice and testing in pronunciation that many English learners in Hong Kong and around the region might not get enough of in the classroom, Mr Chin said.
Most voice-recognition software on the market, including software used for telephone directories and help-lines of companies such as Cathay Pacific, was modelled on adult voices, which were acoustically different from children's, and were designed to recognise whole words, he added.
'This recognition is not recognising it by words - like Cathay is a word, Pacific is a word. It basically breaks a word or some sounds into phonemes, and then it recognises whether this phoneme is correctly spoken, or the second and third phonemes and so on and so forth,' Mr Chin said.
Mr Chin said other projects under development at Astri included software for wireless applications and infrastructure, and projects in photonics. The latter had produced successful prototypes of inexpensive, gigabit-speed equipment that might leave the lab and get commercialised.