Highly evolved speech recognition software
Helping people with speech impediments communicate more effectively can greatly improve the quality of their life, often allowing them to become more independent. Technology has the potential to be a significant driver in this field, as evidenced by the results of the OLP RTD project. One of the major deliverables of the project was the OLP Speech Recogniser - GRIFOS. It was developed at the Universidad Politecnica de Madrid. GRIFOS is an automatic speech recognition system targeted to users with severe speech disorders such as dysarthria. Each patient's particular style of articulation is 'learned' by GRIFOS during a training phase. This is complemented by an existing database containing feedback from other users with the same speech impairment. GRIFOS is used initially to help the patient maintain their speech patterns within certain quantitative thresholds defined by the system. Eventually, the exercises are extended to tackle continuous speech. In the case of dysarthria, GRIFOS is designed to encourage consistency of speech patterns. Like the other OLP software-based tools, GRIFOS embodies a novel approach to speech therapy. It can be a powerful tool not only in the hands of patients, but also their therapists. A prototype is available from the Universidad Politecnica de Madrid and its OLP partners.
