Face up to emotion detection

2 min read

A London-based start-up company has developed technology that reads facial expressions to determine a person’s emotional state.

CrowdEmotion’s facial coding software records individuals’ facial expressions using a camera in a three-step process involving expressions, emotions and actions with potential applications in marketing, sales, education, safety, and gaming.

BBC Worldwide has already deployed the technology in a trial to gauge audience reactions to the BBC’s TV output.

Emotional artificial intelligence will frankly trump humans, within my mind, in the next 10 to 15 years if we have anything to do with it

Matthew Celuszak, CEO of CrowdEmotion

Matthew Celuszak, CEO of CrowdEmotion, told The Engineer that the company aims to become the largest online sample provider and dedicated facial coding provider in the world.

He said: ‘There are two elements to understanding emotions. One is actually capturing the biometric signal like the facial code and the voice, the sweat response and the heart rate et cetera… the bigger question is “What does that mean?”

‘It’s very much a data science problem because emotions are super messy, you need to be able to sift through a lot of very unstructured data to come up with clusters, and that’s what we did.

‘We’re using machine-learning techniques to actually train against and understand what’s clustering.’

The result is a cloud-based facial coding technology that captures the often micro-facial expressions that can be paired with cameras on laptops or smartphones to give an insight into what people are feeling.

Work with BBC Worldwide is providing the corporation with valuable insights into audience engagement, but there are no plans to use the technology to create thresholds or other benchmarks for content.

‘The purpose of this research is for us to understand who each piece of content engages with the most,’ said Emma Finlay, BBC spokeswoman for brands, content & digital. ‘This understanding will help us to serve audiences better by understanding their tastes, which will enable us to be more effective both in getting our content to people who love it, and helping them to discover more.’ 

The trial with BBC Worldwide is also proving CrowdEmotion with insights into the efficacy of its technology whilst simultaneously giving it a large dataset to help it with its stated ambitions..

’Literally in a week I understood how people appreciated BBC content and I had 10,000 face videos,’ he said. ’[That’s] 10,000 people telling me how they appreciated content just by watching it, nothing else. The only impact on them was to watch stuff that hadn’t been screened yet.’

CrowdEmotion took part in Cambridge Wireless Networks’ Discovering Start-Ups competition on October 22, and is a finalist at this year’s LeWeb Start-Up competition being held in Paris, France between 9-11 December.