Project description
Automated evaluation of speech understanding via recording of brain activity
Speech understanding is critical to our ability to communicate, enabling us to respond to people around us, share joy and sorrow, provide services and respond to requests. This vital skill can be compromised by many things, including hearing loss, brain injury, stroke, ageing and other conditions affecting cognitive function. Currently, assessment of speech understanding is done by a clinician and evaluated manually. The EU-funded OSIRIS project is developing an automated system to measure speech understanding that relies on recording brain responses as a person listens, eliminating the need for the person to actively participate. Automation will enable more widespread testing and standardised results; measurement of brain activity will enable the evaluation of persons who may understand what is being said to them but may not be able to express it.
Objective
In medicine, it is often essential to know to what extent a person can understand speech. Current methods to measure speech understanding involve playing specially crafted sentences to the person, letting the person repeat them, and scoring the result manually. This requires active participation of the person, specialised speech material, and is labour intensive. We present a system to measure speech understanding fully automatically. We record brain responses while the person listens to a recording of natural speech, and apply signal processing to derive speech intelligibility. The person does not need to actively participate, which enables testing of a large variety of populations that cannot be tested with traditional methods, such as young children and people with mental impairment. The measurement result relates to everyday communication skills. Our system will make diagnostics cheaper and enable diagnostics outside a specialized clinical center because the system can operate fully automatically.
Fields of science
Programme(s)
Funding Scheme
ERC-POC-LS - ERC Proof of Concept Lump Sum PilotHost institution
3000 Leuven
Belgium