Public announcements in noisy places, such as airports or busy hallways, could be made quieter and clearer, thanks to research from the University of Edinburgh. Scientists there have reportedly developed software that can alter speech before it is broadcast over speakers, making it more audible amid background noise.
“Our findings could offer an alternative, by making speech more intelligible without turning up the volume.”—Dr Cassia Valentini-Botinhao, School of Informatics, Univ of Edinburgh |
In a bid to improve current synthetic voice technology, researchers studied how speech was perceived by listeners. Researchers carried out tests to pinpoint the components of speech that are most easily heard by people in a noisy place. According to them, listeners in loud situations pay most attention to the parts of speech that are easiest to hear, and use those to decipher what is being said.
The researchers developed a mathematical computer program to analyze spoken words and enhance the sounds that help listeners hear what is being said to make speech better understood overall. In tests, the manipulated speech was found to be much easier to understand than natural speech. In some cases, the improvement was the equivalent of lowering noise by 5 dB.
The scientists say that the techniques used in this study could also be used to improve smartphone voices, loudspeaker announcements, or sat-nav systems.
The study, carried out with scientists from Greece, Spain, and Sweden, was funded by the EU, and presented at the recent Interspeech 2013 conference in Lyon. The paper, “Combining perceptually-motivated spectral shaping with loudness and duration modification for intelligibility enhancement of HMM-based synthetic speech in noise,” is authored by Cassia Valentini-Botinhao, Junichi Yamagishi, Simon King, and Yannis Stylianou.
Source: Univ of Edinburgh