NEC accelerates face biometrics 20X with neuroscience-inspired AI engine
NEC Corporation has developed a way to speed up its NeoFace biometrics by up to 20 times while maintaining accuracy rates with a new neuroscience-inspired artificial intelligence technology.
The new AI engine accelerates the speed of applications with real-time analysis of time series data, by accumulating data until it reaches a certain confidence threshold, rather than requiring a particular, previously-fixed amount of data, like typical face biometrics and cyberattack detection engines.
The technology is based on human brain activity when making complex decisions involving sequential accumulation of evidence, and a Sequential Probability Ratio Test (SPRT) technique first proposed in the 1940s and applied to quality control in manufacturing. Neurons in the parietal lobe of the brain’s cerebellar cortex have been found to accumulate evidence in a way reminiscent of SPRT in a recent study, NEC explains.
NEC’s ‘SPRT-based algorithm that Treat As Nth-Order Markov Series’ (SPRT-TANDEM) innovation enables it to overcome the strict prerequisites that have made it difficult to deploy the SPRT in real-world scenarios, the company says.
The same technique can also be applied to cyberattack analysis and other technologies involving time series data to speed them up by similar amounts, and the company will consider further application areas, according to announcement.
NEC presented the technology at the prestigious International Conference on Learning Representations (ICLR) 2021 last week.