Investigation into how novel Quantum Dot Lighting panels can be controlled using artificial intelligence to create therapeutic effects to users

Lead Research Organisation: University of Cambridge
Department Name: Engineering

Abstract

This research will explore the emerging technology of quantum dots (QD) for emitting full colour light in a controlled manner. Full colour QD displays for TV have been introduced by Samsung. they are low power but with unmatched colour palette, definition and luminosity. What the PhD will focus on is taking the knowledge gained from TV displays to lighting. Here plastic panel luminaries incorporating QDs can form ceiling, wall or floor lighting. In essence a room can be a 'light box'. The research question is how do you drive and control 3D lighting with the subjects within the light field. How can we use this light for new functions such as mood enhancement, photo therapy and enhanced work output. In addition the physical functioning of the QD lighting panels must be adjusted to take into account ageing effects. The smart light in this context becomes a 'big data' driven element. The application of artificial intelligence methods to data collected from the lights and user experience to enable and predict new functionalities while optimising the energy usage will be the focus of the PhD research.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509620/1 01/10/2016 30/09/2022
1950511 Studentship EP/N509620/1 01/10/2017 31/03/2021 Chatura Samarakoon
 
Description We evaluated a few neural-network based methods for inferring emotions. Initial findings show that the subjective nature of human affect makes it near impossible to create a general model that can predict a random person's emotional state given the physiological measurements. Thus we needed to carry out user studies to gather information for specific users that we can track over the course of several months of using our wearable device. We created a hardware sensing platform to obtain the physiological measurements with the ability also to sense the colours (light spectra) in the wearer's environment. This data was to be used to find the effect of the spectral content in the ambient lighting environment on a wearer's emotional state.

Although the proposed work was only partially completed, as part of the initial investigation into the perception of colour, we developed a method for picking optimal parameters for approximate colour transforms. Approximate colour transforms is a technique that aims to transform colours in digital images imperceptibly such that they consume less energy to display on an emissive pixel display like OLED.
We conducted a used study through Amazon Mechanical Turk and developed an set of machine learning models that are able to pick optimal transform parameters with a minimum error of nearly 1%. We showed that this colour transform is able to achieve up to 50% energy saving with most participants reporting minimal degradation in image quality.

We are currently investigating the use of optimisation methods to design QD based white lights with a high colour rendering index.
Exploitation Route We aimed to develop a wearable device with a suite of sensors that uses machine learning to infer the wearer's emotional state. This information will then be used to control the QD lighting illuminating the user's environment, such that their emotional state improves. This is done through another machine learning based optimisation loop. The sensors on the wearable device gives access to physiological measurements about the wearer (e.g., heart rate, skin conductance, etc.) and measures the spectral content of the environment they are in.

Future research can be approached using this framework but the approach will need to be validated with user studies.
Sectors Digital/Communication/Information Technologies (including Software),Healthcare