Liopa spun out of Queen’s University Belfast in 2016. The company is onward developing and commercialising research carried out within the university into Visual Speech Recognition (VSR) , effectively automated lip reading.
Liopa have developed LipRead, a AI–based lip reading engine. The company strategy is to have multiple applications across many market segments using the cloud–based LipRead system to perform video–to–text transactions, essentially speech recognition where there is no audio, or where audio exists but is corrupted by background noise. The technology is positioned to increase the user uptake of the Voice–first interface by improving the user experience of voice driven application. There are thus multiple applications for the technology across many disparate market sectors including healthcare, voice assistants, security and online authentication.
Liopa’s technology, LipRead, analyses video of a speaker’s lip movements to decipher speech. Using advanced Machine Learning techniques LipRead analyses video frames to identify discrete lip movements (visemes). Lip movement sequences are then analysed within a pre–trained neural network to produce most likely word/phrase combinations. LipRead can work with standard cameras on today’s smartphones, tablets and PCs. The LipREAD AI–engine is cloud–based and accessible through the use of on–device SDKs.
LipRead can be used to improve the accuracy of today’s speech recognition systems which are all audio–based and degrade in accuracy when used in real–world noisy environments – e.g. using Apple Siri in a busy restaurant. LipRead is audio–noise agnostic and when integrated with audio speech recognition will improve robustness to audio noise.
Additionally LipRead can be used for speech recognition where no audio exists. Use cases here include the ability to analyse CCTV footage to detect whether or not certain words/phrase have been spoken by individuals . From a defense and security perspective this can provide useful context to what is occurring in the footage.
The company was founded by Liam McQuillan (CEO) and Richard McConnell (COO), who co–founded a number of successful NI–based high tech start–ups, and by Dr Darryl Stewart (CSO) and Dr Fabian Campbell–West (CTO) who have been heavily involved in pioneering research in Visual Speech Recognition.Visit website