Skip to main content

Investigating Speech Processing in Realistic Environments

Article Category

Article available in the folowing languages:

Help for the hearing-impaired

An EU team studied the various ways that individuals interpret speech under noisy conditions. The work yielded an ultrasound tool to help with such assessment, plus an improvement to speech recognition systems.


Most people can interpret speech under various combinations of suboptimal conditions. Also, the high individual variability in coping mechanisms suggests ways of tailoring hearing aids and other audio devices to individual users and changing circumstances. The EU-funded INSPIRE (Investigating speech processing in realistic environments) project's research drew together factors affecting speech intelligibility. The consortium also ran a training programme. The team first specified the key mechanisms by which listeners' perceptions and interpretations change during realistic listening situations. Researchers also developed an ultrasound device as a first step towards technological applications responsive to degraded noise conditions. Next, the group detailed the ways in which listeners vary depending on language, hearing ability and memory, plus how listeners respond differently to real-world speech. The team developed laboratory speech tests able to simulate parts of real-world communication and reliably reproduce the full range of individual responses. Researchers modelled how the effects of age and hearing loss affect listeners' decoding ability, and to what extent hearing aids restore realistic perception. The team studied individual differences in sensitivity to masking. INSPIRE developed models able to predict misperceptions in various suboptimal circumstances. The misperceptions were analysed to find the signal properties that caused confusion. Results led to refinement of current models. A further stage improved the robustness of current automatic speech recognition systems. The development included a superior speech-noise separation algorithm. Additionally, INSPIRE managed a comprehensive training programme, including workshops, keystone skills courses and complementary skills. The work yielded over 50 journal papers, and helped create a stable research infrastructure. The infrastructure included complementary research projects such as BigListen, speech stimuli synthesis software and a platform to advance auditory modelling. INSPIRE's work may lead to technologies enabling individualised use of hearing-assistance and other audio devices. Such an outcome would benefit audiological research and the hearing-impaired population.


Hearing-impaired, audio devices, INSPIRE, speech processing, realistic environments

Discover other articles in the same domain of application