As many users in developing regions are literate or semi-literate, spoken language input and output plays a major part in the design of user interfaces to deliver ICTs. To that end, we investigate the problem of providing speech interfaces as alternative or as complementary, to visual interfaces.
Within speech recognition, we are working projects to build tools and applications using speech recognition for Indian languages, designing low-power and low-cost hardware for speech recognition.
This project aims to develop low-cost, low power solutions in custom hardware for speech recognition. Even though reliable recognition of large-vocabulary fluent speech is still not fully achievable in hardware at low-power, providing highly accurate recognition for isolated word or limited vocabulary systems using cheap hardware is an achievable goal. We want to design new user-interfacing solutions, suited for accomodating a variety of users, and deployable on cheap, low-cost and power constrained devices. Providing inexpensive speech recognition may also alleviate the need for costly and energy-cosuming LCD interfaces.
For more details, read our paper, Hardware Speech Recognition on Low-Cost and Low-Power Devices in DAC, 2005 or contact Sergiu Nedevschi and Rabin Patra .
The Open Sesame project builds speech tools for building small-vocabulary spoken dialog systems. The systems are powered on rapidly-deployable speech recognition in a variety of language and dialect contexts. Primary work on this has been targeted towards agricultural kiosks in Tamil Nadu, India.
The Open Spell project grew from a desire to provide local language content to children and adults in developing regions where access to computers are growing, but for which the development of software created in local languages continues to lag. In many of these regions, multiple languages are actively spoken, some with more social prestige than others. Our project sought to provide a multilingual game, which would allow users to play learning games in a variety of languages.
Some publications are here