It has been reported that PaeLife, a collaborative international project designed to aid elderly people in assisted living, has made significant improvements in speech recognition technology for elderly people in various European languages, following a study involving over 3,000 participants.
Since the speech of elderly people tends to be slower and quieter than the speech of most young people, elderly people are often at a disadvantage when working with audio-operated technology designed for younger, clearer voices.
PaeLife selected speakers from Universities of the Third Age, various care institutions, associations and social clubs for the elderly from across Poland, Hungary and France. Previously, data for Portuguese speakers had been collected and transcribed by the Living Usability Lab and the Smart-Phones for Seniors projects.
To ensure a variety of accents, the 3,000 speakers were selected from different regions of the four countries. In total, over 640 hours of speech recordings were collected in the four languages, thus ensuring that the process of optimising elderly speech recognition software could proceed to the next stage.
One of the main goals for the PaeLife project is to make the software understand what is being said without misinterpreting false-starts, mispronounced words, pauses, coughing and the sound of doors opening and closing. To counteract this, the recordings were marked in the transcribing process to research further how these issues could interfere with understanding what is being said.
Harry Phipps and Esther Galdo
Hämäläinen, Annika, António Teixeirac, Nuno Almeidac, Hugo Meinedoa, TiborFegyó, Miguel Sales Diasa (2015) Multilingual speech recognition for the elderly: The AALFred personal life assistant. Procedia Computer Science 67: 283-292.