Projecto ARIA
PROJECT ARIA - Ambient-assisted Reading Interfaces for the Ageing-society
Description
The novelty of this project task is the target population to be addressed: senior or elderly citizens. 
It is well known that elderly people do appreciate story-telling. Aging humans do lose reading capabilities, mainly due to their loss of vision. 
“New technologies” can come to their rescue but, unfortunately for many of them, they can be also became “a form of discomfort”. 
So, easier and more intuitive ways to interact with computers are welcome by them. Speech interaction is the most natural way for older adults to interact, 
so it is the best bet to allow the elderly population to adopt the use of computers for their activities, and in particular to let them return to their reading habits. 
Verbal commands could be uttered in order to navigate in an ambient assisted reading environment thereby simplifying the interaction with computers, 
avoiding unfamiliar keyboards. Although isolated word recognizers are quite common and well known, their usage by senior citizens is not widespread. 
We propose to work with a necessarily small number of senior citizens in order to get some insight on the problems that may show up in ASR. We intend to deal with these 
problems by evaluating two different approaches: speaker adaptation and training new word models. In the former, we intend to assess the ASR performance of short 
navigation/browsing commands using phone models obtained by a retraining process of already available (speaker independent) phone models. This approach, being 
computationally demanding, has the advantage of being scalable, since new words or commands not existing in the database can easily be included. In this case, all the 
command words to be used in the navigation tasks must be previously recorded. Should new or additional words be required, new recordings must be done in order to train 
those new occurrences. Expected results will be a working speech recognizer, with a set of models tailored for older adults, allowing for the recognition of a limited set 
of document navigation and information retrieval commands. So, the purpose of this research grant is to develop and test a browser interfacing a Speech Recognition Engine 
and also the framework for testing and evaluation.

Type of Position: Research Fellowship (Bolsa de Investigação)
Durations: 6 Months
Limit Date to Reply: 14-Feb-2011
Contact: António Joaquim dos Santos Romão Serralheiro
Scholarship Email: antonio.serralheiro@inesc-id.pt